**Section I: Empirical Research on Congregations and Denominational Variations**

### **Religion and Infant Mortality in the U.S.: A Preliminary Study of Denominational Variations**

### **John P. Bartkowski, Xiaohe Xu and Ginny E. Garcia**

**Abstract:** Prior research has identified a number of antecedents to infant mortality, but has been focused on either structural (demographic) forces or medical (public health) factors, both of which ignore potential cultural influences. Our study introduces a cultural model for explaining variations in infant mortality, one focused on the role of community-level religious factors. A key impetus for our study is well-established religious variations in adult mortality at the community level. Seeking to extend the growing body of research on contextual-level effects of religion, this study examines the impact of religious ecology (*i.e.*, the institutional market share of particular denominational traditions) on county-level infant mortality in the U.S. Analyses of congregational census and Kids Count data reveal that a high prevalence of Catholic and most types of conservative Protestant churches are associated with lower rates of infant mortality when compared with counties that feature fewer Catholic and conservative Protestant congregations. However, communities with a large proportion of Pentecostal churches exhibit significantly higher infant mortality rates. After discussing the implications of these findings, we specify various directions for future research.

Reprinted from *Religions*. Cite as: Bartkowski, J.P.; Xu, X.; Garcia, G.E. Religion and Infant Mortality in the U.S.: A Preliminary Study of Denominational Variations. *Religions* **2011**, *2*, 264–276.

### **Introduction**

Infant mortality, or the number of infant deaths among children one year-old and younger per 1,000 live births, has long been recognized as a preeminent indicator of the overall health among nations [1,2]. As a cross-cultural and historical marker of social development, the infant mortality rate (or IMR) is a critical component of quality of life indices. Such indices are regularly used by international and domestic governmental agencies, as well as prominent nonprofits concerned about the welfare of children, to rank nations, regions, and states in terms of their social development and their dedication to the well-being of their most vulnerable citizens.

The United States presents a vexing dilemma with respect to infant mortality. Although the U.S. maintains a remarkably high per capita income, its IMRs are disproportionately elevated when compared with other developed nations in the West [1,3]. The infant mortality paradox in the United States continues to generate concern among public health officials and policymakers, many of whom have called for immediate action to reduce the unusually high IMRs in the U.S. [1,4].

Rather alarmingly, the most recent efforts to reduce U.S. IMRs have yielded less than stellar results. As part of its *Healthy People 2010* initiative, the U.S. had been aiming to reduce its IMR to 4.5 infant deaths per 1,000 live births [5]. However, most of the first decade of the twenty-first century came and went with little significant change in the U.S. infant mortality rate. The U.S. IMR was 6.89 infant deaths per 1,000 live births in 2000 and had dropped only slightly to 6.86 in 2005, and such stagnant results have not been observed since the 1950s [1]. Quite tellingly, the U.S. has now revised its 2020 objectives to aim for an IMR of 6.0 per 1,000 live births [6].

This study explores the potential effects of religious ecology (that is, community-level denominational market share) on this important social indicator. Why might community-level religiosity be expected to influence the infant mortality rates observed across U.S. counties? Perhaps most notably, the respective market shares of various faith traditions have been shown to influence adult mortality and morbidity patterns across the U.S. [7,8]. Moreover, religious institutions often act as advocates for the well-being of families and children [9,10], such that family ministry programs play a central role in the work of many congregations [11]. Given such prior research, we surmise that religious ecology might influence county-level infant mortality rates in the U.S.

We begin our investigation of this phenomenon with a review of the literature on infant mortality. We then proceed to expand on our rationale for examining how religious ecology might influence infant mortality, arguing for a cultural model to explain IMRs net of structural factors. Next, we describe the population-based sources of data used for the this study, including Kids Count Data, the Glenmary Census of Churches, and select U.S. Census data from the year 2000. After reviewing the results of our investigation, we conclude by considering the implications and limitations of our study while also specifying directions for future research.

### **Infant Mortality: Summary of Key Antecedents**

Scholars have long examined health disparities across social groups (communities, states, nations) [12-17], and the study of infant mortality has extended this line of research to explain differentials in live birthrates. One avenue of research pursued by demographers studying this issue has explored cross-national patterns in infant mortality [3,14,18-19]. In general, these studies have linked high rates of infant mortality to differential access to socioeconomic resources, unsanitary living conditions in developing countries, and restricted health care access [3,19].

Another line of research, more germane to our investigation, has focused on infant mortality differentials within the United States [3,20-21]. Scholars focused on the U.S. have pinpointed three community-level factors associated with inordinately high infant mortality rates. First, and not surprisingly, infant mortality is highly correlated with social inequality, primarily in the form of income and racial stratification [3,22-25]. Concentrated socioeconomic disadvantage has been well documented to create a standard of living that threatens the welfare of families and the well-being of children [26]. Poverty constricts resources that might otherwise be used to promote preventive and, quite significantly in this case, prenatal health care. Within the U.S., infant mortality rates are significantly higher in regions, states, and counties marked by concentrated socioeconomic disadvantage.

In addition, infant mortality varies by locale [24,27]. Although one might expect that rural areas would have higher infant mortality rates due to the lack of a health care infrastructure, metropolitan areas typically exhibit higher rates of infant mortality [27]. This pattern is explained by the concentrated poverty (compromised health facilities, overcrowded housing, and disadvantaged neighborhoods) that marks many urban core areas in the United States [28,29].

Finally, and not unrelated to the foregoing point, health care access has also been linked to infant mortality [3,15,24]. Infant mortality tends to be significantly greater in communities marked by a lower number of hospital beds and physicians per capita. While the links between health care inaccessibility and higher infant mortality rates are again intuitive, there are several possible mechanisms at work here. The provision of preventive prenatal and infant care may be limited by a restricted health care infrastructure. Moreover, when health complications arise during deliveries and postpartum, the ability for successful medical intervention may be hampered in communities lacking a sufficient health care infrastructure.

### **Religious Ecology and Infant Mortality: Toward a Cultural Model**

What is religiosity? To most Americans and many scholars, religiosity is conceptualized in terms of individual beliefs, convictions, and practices. In this sense, religiosity is commonly defined in terms of a person's beliefs about God or sacred scripture (e.g., images of God as authoritative father *vs.* loving friend, views about the inerrancy of the Bible), the subjective importance of religion (e.g., religious salience with respect to major life decisions), or an individual's religious practices (e.g., frequency of attendance at weekly worship services, affiliation with a particular denomination). This study is founded on the insight that religion is not solely an individual attribute. It can also function as a group property that influences the climate and quality of life in a community.

Taking a cue from previous research, we use the term "religious ecology" to examine the community-level character and influence of religion [7,30-33]. Religious ecology can be defined in a number different ways. Perhaps the crudest measure of religious ecology is the raw number of congregations (e.g., churches, synagogues, and mosques) or congregants in a given community, adjusted for the population of that community. More sophisticated measures of religious ecology examine the "market share" of different types of congregations (e.g., conservative Protestant *vs.* mainline Protestant *vs.* Catholic), the prevalence of civically engaged denominations (*i.e.*, a combination of religious groups that are outreach-oriented), and even the presence of faith-based organizations (e.g., religious nonprofit social service agencies) in a community, again with respect to the size of the local population.

The review of prior scholarship featured above demonstrates that, to this point, scholars have utilized either demographic or public health models to explain infant mortality rates. In explaining IMR differentials, demographic models analyze the influence of socioeconomic disadvantage while public health scholars examine gaps in health care provision and access. By pointing to the possible influence of religious congregations on IMRs, we argue here for a cultural model. This cultural model recognizes that religious congregations can influence collective norms whose force is evident in contextual-level outcomes, including health outcomes such as IMRs. It is this insight that drives a great deal of the research on the role of religion in forming moral communities [31,33]. Congregations can create a moral ethos in communities that, in a very Durkheimian fashion, yields distinctive social outcomes, including those related to community-level morbidity and mortality patterns.

Beyond the complement that a cultural model might offer to existing approaches, why might community-level religiosity influence infant mortality? Three possible reasons are immediately apparent. First, a great deal of research has demonstrated a clear linkage between religion and health. At an individual level, religious involvement is inversely associated with morbidity and mortality [34-40]. Scholars who have observed this relationship have argued that religious people adhere to healthier lifestyles and practices, experience enhanced social support, and have greater coping resources than their non-religious counterparts [41,42].

The robustness of such findings notwithstanding, survey-based studies on religious differences in individual health are insufficient to warrant an ecological investigation of religion and infant mortality. As demographers are quick to point out, patterns that are observed at one level of analysis (*i.e.*, among individuals) cannot be presumed to operate identically at another (*i.e.*, among whole populations). So, what additional evidence is there that might justify an ecological investigation of religion and infant mortality?

As it turns out, community-level religiosity has been linked to collective mortality patterns (population-based death rates). A recent study by Blanchard and colleagues [7] documented that mortality rates were significantly higher in conservative Protestant communities than in counties dominated by other faith traditions such as mainline Protestantism and Catholicism. Blanchard *et al.* explained these differences by arguing that the otherworldly, anti-institutional characteristics of conservative Protestantism led to a diminished investment in community infrastructures that promote this-worldly pursuits such as health care. Interestingly, Blanchard and colleagues also performed decomposition analyses to examine mortality rates across communities characterized by different types of conservative Protestant churches, namely, fundamentalist, Pentecostal, evangelical, and other conservative Protestant congregations, with the last of these a residual category for Bible churches not clearly situated in any of the first three subgroups. Blanchard and colleagues found that communities characterized by a larger number of fundamentalist and Pentecostal churches had higher mortality rates than those dominated by their evangelical and other conservative Protestant cousins. They explained these findings by arguing that although evangelical and Bible churches are careful to distinguish themselves from the secular world, their desire to attract converts entails maintaining a degree of engagement with the surrounding culture. It is this same secular culture that is more thoroughly shunned by their fundamentalist and Pentecostal counterparts. The especially high rates of mortality evident in Pentecostal communities are likely also a product of the centrality of faith healing within this subgroup and a concomitant distrust of conventional medicine.

Beyond the literatures on religion, health, and mortality, there is also plenty of scholarship underscoring the potential benefits of religion on family life. A primary focus of religious congregations is the provision of services to families, youth, and young children [9-11]. This scholarship underscores the "pro-family" character of religion, with congregations prioritizing family ministry programs over much of the other work they undertake. Moreover, recent research has revealed that strong linkages between religious institutions and families can be beneficial for child development [10]. Taken together, these bodies of research demonstrate that attention needs to be given to the influence of cultural factors, such as religion, on infant mortality.

In light of the foregoing research, and particularly scholarship on denominational market share and adult mortality patterns, we offer the following hypotheses about religious ecology and infant mortality.

H1: Counties with a greater proportion of Catholic and mainline Protestant congregations will exhibit lower infant mortality rates, while those with a greater proportion of conservative Protestant congregations will exhibit significantly higher infant mortality rates.

H2: Among conservative Protestant faith traditions, counties with a greater proportion of fundamentalist and Pentecostal congregations will exhibit significantly higher infant mortality rates, while those with a greater proportion of evangelical and other conservative Protestant congregations will have significantly lower infant mortality rates.

### **Data and Methods**

The data enlisted in this study are derived from three different sources. Kids Count data were used to generated our dependent variable, namely, infant mortality rates. These rates are available through Kids Count, and were not calculated by the authors. Where possible, year 2000 Kids Count data were used to construct this variable. It is worth noting that Kids Count data provide infant mortality rates for a restricted number of counties, namely, those in which at least one such incident occurred during a given year and those which reported infant mortality data to the federal government. No incident counties and unreported data counties reduce the number of counties available for analysis and thereby compromise these data somewhat. Given clusters of county-level case attrition in the Mountain West, we supplemented infant mortality data for three states (Montana, New Mexico, and Wyoming) through an interpolation method. Missing year 2000 data were interpolated for 56 counties in Montana, 33 counties in New Mexico, and 23 counties in Wyoming by using an average of infant mortality rates from later years (ranging from 2001 through 2005). The combination of available data and interpolated data produced a study region of 1,900 counties. The 112 counties for which data were interpolated do not threaten the validity of our study because these counties constitute a small proportion of our sample (5.89 percent of all counties in our study region).

Our primary independent variables reflecting county-level religious ecologies, the Glenmary Census of Churches (2000), were retrieved from the Associated Religion Data Archive. Religious denominations in this dataset were coded consistent with the framework developed and utilized in Blanchard *et al.* [7]. First, major faith traditions were coded into four categories: conservative Protestant, mainline Protestant, Catholic, and other. A series of denominational variables were then created to reflect the number of congregations per 1,000 residents for each denominational family in a county. (Standardizing this measure as the number of denomination-specific congregations per 1,000 county residents creates a comparable baseline of comparison across counties of different sizes.) This coding scheme allows for aggregate analysis of the effects of religious ecology on infant mortality for broad denominational families. In our analysis, each denominational tradition is treated as a continuous variable, such that a specific change (standard deviation increase or decrease) for the denominational variable is associated with a specific change (standard deviation increase or decrease) for the infant mortality rate. When discussing the results of these analyses, we report standardized regression coefficients. Second, to conduct our decomposition analyses, we recoded the conservative Protestant category into four subcategories: fundamentalist, evangelical, Pentecostal, and other conservative Protestant (the last category serving as a residual category that did not fit into the first three categories). The analytical strategy used in this phase of the investigation (that is, unit change in the denominational variable compared with unit change in the infant mortality rate variable) is the same that was used for the major faith traditions. Here again, denominational families are treated as continuous variables.

In light of the previous literature on this subject, we control for three key ecological factors known to influence infant mortality rates. To control for concentrated disadvantage and account for the connection between poverty and race-ethnicity in American society, we include the following covariates (generated through 2000 county-level Census data) in our regression models: percent of population under 18 living below the poverty line, and the percent of the population under age 18 that is black. Given spatial variations in infant mortality, we control for region of the country: Northeast, Midwest, West, and South, with the last of these serving as the reference category. We use ordinary least squares (OLS) regression to conduct these analyses.

### **Results**

The results of our analyses are reported in Table 1. As can be observed from the table, our study captures nearly two-thirds of the counties in the U.S. Therefore, we use a study sample of 1,900 U.S. counties (from a total of 3,143) to conduct both aggregate analyses (major denominational groupings, Model 1) and decomposition analyses (conservative Protestant subgroups, Model 2).

Recall that Hypothesis 1 predicted lower infant mortality rates for counties with a greater proportion of Catholic and mainline Protestant congregations coupled with higher infant mortality rates for conservative Protestant counties. Our results in Model 1 lend only partial support to this hypothesis. Counties with a high proportion of Catholic congregations have a significantly lower level of infant mortality than counties with fewer Catholic churches. This finding is consistent with our hypothesized effects. However, there is no effect for mainline Protestant counties, which were expected (like Catholic counties) to have a significantly lower infant mortality rate. Moreover, although we expected conservative Protestant counties to exhibit a significantly higher infant mortality rate, the findings in Model 1 do not support this hypothesis. The conservative Protestant variable is not statistically significant.

Next, we turn to our decomposition analyses, which estimate the net effects of our conservative Protestant subgroup variables. Recall that we hypothesized that counties with a greater proportion of fundamentalist and Pentecostal congregations would have significantly higher infant mortality rates, while evangelical and other conservative Protestant counties were expected to exhibit significantly lower rates. These findings, we expected, would mirror religious influences on adult mortality.


**Table 1.** Standardized Ordinary Least Squares (OLS) Regression Estimates Predicting Infant Mortality, n = 1,900 counties.

Note: The following variables are controlled: percent of population under age 18 who are below poverty (2000); percent of population under age 18 that is black (2000); and region (Northeast, Midwest, West, and South). Dependent variable: infant deaths per 1,000 live births.

As the coefficients in Model 2 of our table indicate, this hypothesis is strongly supported. Based on previous findings related to adult mortality, we expected significant inverse effects for evangelical and other conservative Protestant counties. These expectations were generally met. Counties with a high proportion of evangelical congregations exhibit a significantly lower infant mortality rate, as do counties with a higher proportion of other conservative Protestant congregations. The effects for the Pentecostal variable also support our hypothesis, given that we expected significantly higher infant mortality rates among Pentecostal counties. Our expectations regarding effects for counties with a high proportion of other conservative Protestant congregations were also substantiated. Counties with a large proportion of other conservative Protestant congregations (e.g., Bible churches) exhibit a significantly lower infant mortality rate when compared with counties that have relatively few of these types of congregations. The only finding that runs contrary to our expectations pertains to counties with a high proportion of fundamentalist congregations, which exhibit a significantly lower infant mortality rate. Comparing Models 1 and 2, we observe persistently significant effects for the Catholic variable, which remains significant in the decomposition analyses (Model 2). Contrary to our initial expectations, counties with higher proportions of mainline Protestant churches are more likely to exhibit high infant mortality rates in Model 2, though this relationship was insignificant in Model 1.

### **Discussion**

This study examined the effects of religious ecology (that is, community-level denominational market share) on infant mortality. We argued that community-level religiosity may influence infant mortality rates in U.S. counties because (1) adult mortality is affected by the social ecology of religion in communities, and (2) congregations and denominations position themselves as pro-family institutions. Several significant findings emerged. In most of our regression models, higher levels of community religiosity were linked to lower infant mortality rates. However, this pattern was not uniformly observed for all denominational families. In what follows, we summarize and explain our findings. Thereafter, we highlight limitations of our study and identify some promising directions for future research.

Consistent with patterns for adult mortality [7], we found that counties with a higher proportion of Catholic congregations were characterized by lower infant mortality rates. This pattern can be explained by many of the same arguments enlisted by Blanchard and colleagues [7] in which civically engaged denominations can be distinguished from their civically insular counterparts [31,33,43]. Catholic congregations are civically minded, externally oriented institutions that place a premium on the creation of community-level care. This argument is consistent with the Catholic theological commitment to "subsidiarity," that is, caring for persons not just as individuals but as groups situated within communities. Contrary to the findings of Blanchard and colleagues, there were inconsistent effects observed for counties characterized by a high proportion of mainline Protestant congregations. Although this pattern is difficult to explain in any definitive fashion, it is possible that the especially strong Catholic commitment to subsidiarity distinguishes Catholic counties from their mainline Protestant counterparts.

Interestingly, the aggregate analyses examining overall conservative Protestant effects on infant mortality did not meet our expectations. Thus, while conservative Protestant counties have produced significantly higher rates of adult mortality, no effects surfaced in such counties where infant mortality is concerned. This non-finding was followed by decomposition analyses that estimated the effects of various conservative Protestant subgroups on infant mortality. These subgroups included fundamentalists, evangelicals, Pentecostals, and other conservative Protestant congregations. Decomposition analyses could reveal subgroup variations that might be masked among conservative Protestants at large. And, indeed, our decomposition analyses did just that.

The decomposition analyses revealed lower rates of infant mortality in counties with higher proportions of fundamentalist, evangelical, and other conservative Protestant congregations. Previous findings on community-level religious variations in adult mortality [7] led us to expect that evangelical and other conservative Protestant counties would exhibit lower infant mortality rates, while fundamentalist and Pentecostal counties would exhibit higher infant mortality rates. Thus, the key difference between infant mortality and adult mortality, where religious effects are concerned, is found in counties that feature a large proportion of fundamentalist churches. Why would three of the four conservative Protestant subgroups (that is, all of them except Pentecostals) create community climates that are less conducive to infant mortality? Conservative Christian churches are not just generic pro-family institutions. They are also pronatalist, and are especially vigorous at exhibiting what they argue is a "defense" of those who cannot protect themselves, namely, the young. Nowhere is this position more evident than in their opposition to abortion. It is quite possible that the pro-life stance taken by many conservative Protestant congregations make their communities particularly attentive to threats to the welfare of the young. The rhetoric and programs in these congregations may privilege the well-being of the young in a way that is quite different from the manner in which adult well-being is treated. In the individualistic worldview of conservative Protestantism, adults may be expected to take care of themselves in a way that children could not be expected to do so.

Why, then, do counties with a higher proportion of Pentecostal congregations run counter to this pattern that is so evident among the three other subgroups of conservative Protestants? One of the key elements of the Pentecostal faith tradition is a commitment to faith healing [44]. It is possible that in such communities, a collective wariness toward medical interventions leads to an ethos in which prenatal or postpartum care is institutionally and normatively deemphasized. Perhaps both preventive care and medical intervention in the face of complications are collectively defined as a demonstrated lack of faith in God. Although more research is clearly needed on this front, the much higher rates of adult mortality (previous research) [7] and infant mortality (our study) in counties with higher proportions of Pentecostal congregations lend credibility to this interpretation.

This study is not without limitations. First, we cannot establish direct causal connections between religious ecology and infant mortality in this study. In a preliminary investigation of this sort, we have included only select control variables. Future research could control for a wider array of factors, such as health care access, to determine if such factors mediate or moderate the relationship between religious ecology and infant mortality. Moreover, in this cross-sectional investigation, we cannot presume the direction of causality. We view it as reasonable to presume that religious ecology exerts an influence on infant mortality, but cannot dismiss arguments about reverse causality without additional data. It is possible that communities with lower rates of infant mortality are more receptive to religious institutions because the faith of their residents is not "tested" in the same way as communities marked by higher infant mortality rates. Consequently, a longitudinal analysis using 2000 and 2010 data presents itself as a promising direction for future research. Such a follow-up investigation would be valuable for introducing more contemporary data, but also for trying to determine possible causal connections between religious ecology and infant mortality. That investigation would require different methods and statistical techniques than we have used here (e.g., controlling for changes in religious ecology over time) and, as such, is beyond the scope of our current investigation.

Second, as mentioned in our methodology section, the dependent variable for this study was drawn from Kids Count data and, therefore, reflects a restricted sample of counties in the U.S. Some of these counties were characterized by no incidents and others simply seemed not to report data on this vital statistic. Therefore, while the findings presented here are the product of rigorous analytical procedures and statistical tests, these data limitations lead us to call for more research on religious variations in infant mortality. Given the fact that our study is predicated on a restricted county sample, the effort undertaken here must be treated as a preliminary investigation into the phenomenon. Additional research is needed with the full universe of U.S. counties to determine if the findings generated with our more restrictive sample of counties hold across all U.S. counties.

Third, as illustrated by our literature review, a good deal of current research on infant mortality adopts a comparative perspective by exploring cross-national patterns and trends. Our study was limited inasmuch as it focused on infant mortality in one country, namely, the United States. Additional research is needed to explore the ecological effects of religion on infant mortality across national borders. There is much to gain from broadening the investigative lens with cross-national comparisons, as there may be peculiarities associated with particular traditions in the U.S. context that may not be observed elsewhere in the world. For example, the Catholic penchant for community engagement and its association with lower rates of infant mortality may be due to the minority status and historic marginalization of Catholicism in the United States, a predominantly Protestant country.

Fourth, there are some fruitful alternative means of defining religious ecology that we did not explore here. For example, previous research has demonstrated that a preponderance of civically engaged denominations, as initially operationalized by Tolbert and colleagues [43], often produce salutary community outcomes [31,43]. The influence of these types of denominations on infant mortality is quite worthy of investigation. In addition, infant mortality rates could vary in terms of the religious diversity exhibited in local communities. Because religious competition and commitment may be greater in communities characterized by a high degree of denominational diversity, the use of an index of religious dissimilarity presents an intriguing prospect for further specification of the relationship between religious ecology and infant mortality. Once again, this last avenue of inquiry could be quite fruitfully examined not only with U.S. data, but with international data as well.

These limitations and promising directions for future research notwithstanding, our investigation adds significantly to the literature on infant mortality. To this point, explanations of infant mortality differentials have been dominated by either demographic or public health approaches. Demographers have enlisted a structural model that emphasizes, among other factors, how poverty and concentrated disadvantage contribute to higher infant mortality rates. By contrast, public health approaches utilize a medical model to explain infant mortality. The medical model examines how factors such as health care access (e.g., physicians per capita) may contribute to infant mortality rates. This study represents a first step toward examining a cultural model of infant mortality. Central to this cultural model is the moral ethos that religious institutions can create in communities, and the way in which this ethos can produce real-world effects on population health, in this case, infant mortality. While there is much additional work to be conducted on this topic, our study demonstrates that cultural factors should no longer be ignored in exploring the determinants of infant mortality in the U.S.

### **Conclusions**

This study revealed that the social ecology of religion (denominational market share) is associated with county-level infant mortality rates. Using data from the year 2000, we found that counties with a high proportion of Catholic churches are significantly more likely to have a low infant mortality rate. This finding is best explained by the emphasis that Catholicism places on creating a vibrant civic infrastructure, particularly one focused on promoting population health and well-being. Although our general measure of conservative Protestant market share did not produce any significant effects in a preliminary model, our follow-up decomposition analyses compared the respective influences of different types of conservative Protestant congregations (fundamentalist *vs.* evangelical *vs.* Pentecostal) on county-level infant mortality rates. The decomposition analyses demonstrated that counties with a high proportion of fundamentalist, evangelical, and other conservative Protestant congregations (Bible churches) have significantly lower infant mortality rates, while those with a high proportion of Pentecostal churches have significantly higher infant mortality rates. It is quite likely that the pronatalist tendencies of fundamentalism and evangelicalism (advocacy for children and the unborn) contribute to significantly lower infant mortality rates in areas where these churches enjoy a large market share. Pentecostalism presents an interesting deviation from this pattern. We suspect that Pentecostal suspicion of conventional medicine and its reliance instead on faith healing accounts for the higher infant mortality rates in counties with many of these congregations. Our study is, of course, a preliminary investigation of this phenomenon. However, these findings meaningfully extend previous research on religion and health while suggesting future opportunities that are ripe for investigation.

### **References**


### **Bringing the Congregations Back in: Religious Markets, Congregational Density, and American Religious Participation**

### **Troy Blanchard, Samuel Stroope and Charles Tolbert**

**Abstract:** We draw on the organizational ecology tradition to frame the relationship between the religious environment of a community and local religious participation. Prior research linking religious environments to religious participation downplays a key organizational aspect of religion: the congregation. Following the organizational ecology usage of density, we argue that *congregational density*—the number of congregations per person within a community—impacts religious involvement by providing opportunities for participation and by fostering social accountability networks within congregations. Drawing on data from the 2000 Social Capital Community Benchmark Survey, we test the hypothesis that congregational density in a locality is associated with greater religious participation by residents. Our findings indicate that persons residing in congregationally dense communities are more likely to be members of churches, to attend church regularly, to participate in church-based activities, to participate in non-church religious organizations, to volunteer for religious work, and to give to religious causes. These findings hold while controlling for an array of individual and contextual-level variables. This notion of congregational density suggests that local factors transcend broader theological and/or denominational boundaries, resulting in variations in religious participation and commitment.

Reprinted from *Religions*. Cite as: Blanchard, T.; Stroope, S.; Tolbert, C. Bringing the Congregations Back in: Religious Markets, Congregational Density, and American Religious Participation. *Religions* **2014**, *5*, 929–947.

### **1. Introduction**

Sociologists have a longstanding interest in identifying the determinants of American religious participation. For some researchers, variation in religious participation is explained by individual-level demographic and socioeconomic traits. Others have focused on how the relationship between those individual-level traits and religious participation may vary across communities. Still other researchers have studied how religious participation and other social and economic outcomes are linked to the local religious ecology as defined by the presence of specific denominations. Some researchers have conceptualized the religious ecology as the mix of religious denominations present in the community and the extent of interdenominational competition. In virtually every case, the local religious ecology is conceptualized and measured in terms of religious denominations.

These various approaches have proven valuable in explaining the role of the local religious ecology on community welfare. However, these conceptualizations of the religious ecology neglect the local congregational population that may mediate the relationship between a denomination's theological program and the individual religious adherent. To be sure, prior studies have devoted a great deal of effort to understand how congregations work, why some differ from others, and how congregations may interact with the broader community. Though, only rarely have researchers investigated the effect of the local religious ecology's *congregational population* on religious participation. This omission highlights an important research question: why are congregations overshadowed by denominations in macro-level explanations of religious commitment and participation?

Drawing on organizational ecology theory, we develop an explanation of how the local population of congregations influences participation in religious activities. Instead of focusing exclusively on denominations, we consider the embeddedness of religious participation within a local congregational population. Our central argument is that *congregational density*—the number of congregations relative to the local population—impacts the relative size and structure of congregations in a community. We hypothesize that a dense population of congregations will increase the likelihood of participation among community members. In contrast, people living in communities with a small number of churches relative to the local population will be less likely to participate in religious activities.

Our analyses test for the effect of congregational density on individual-level religious participation controlling for personal attributes known to affect the likelihood of religious participation such as age, gender, religious affiliation, and socioeconomic status. We employ several measures of individuals' religious participation: church membership, service attendance, participating in church activities, nonchurch-based religious organizational membership, volunteering for religious organizations, and giving to religious entities. We also specify potentially confounding local market factors, such as socioeconomic disadvantage. The models are also specified to minimize model endogeneity that can pose a challenge to analyses such as this one. It could be argued that highly religious areas may produce more churches. We posit just the opposite, but recognize that causal direction is not a trivial issue. We will return to a discussion of it after we develop our research hypotheses below.

### **2. Rethinking the Religious Environment: From Denominations to Congregations**

The study of religious markets and religious participation finds its roots in Durkheim's assertion that religion is a primary integrating force that engenders social solidarity [1]. Berger [2] extends Durkheim's perspective by arguing that markets where all community members adhere to a single theological orientation yield a greater level of religious participation and social integration. More recently, Stark, Finke, Iannaccone, and others [3,4] have provided evidence that a wide variety of religious choices within a community (*i.e.*, religious pluralism), rather than a single denomination, increases rates of religious participation. This approach is based on three key propositions: (1) unregulated religious economies will tend to be pluralistic; (2) pluralistic religious environments engender firm specialization; and (3) specialization generates religious participation. Thus, the lack of regulation in the religious marketplace permits the development of a wide array of faith traditions that become tailored to specific population segments [5,6].

Within the religious pluralism literature, the religious economy is organized by denominations—what Stark and others refer to as religious "firms" [3]. Although the pluralism literature contains occasional references to congregations and church leaders as actors in the history of the U.S. religious marketplace [6], religious congregations are largely absent from the empirical measurement of the religious economy. The degree of religious pluralism is measured using an index of denominational market concentration, the Hirschman-Herfindahl Index, which summarizes the concentration and/or dispersion of adherents across denominations in a given community. In this way, the religious establishment is equated with a denomination. Implicitly, congregations within a local religious market area are assumed to be uniform representatives of a denomination. Left veiled by such a measure is the variation of congregations within denominations across space.

Just as business establishments vary within parent corporate enterprises, congregations vary within denominations in important ways. First, regardless of denominational affiliation, congregations are socially embedded in communities that shape congregational life. Prior studies highlight the importance of the culture and social context of a locality in shaping a congregation [7,8]. Catholic congregations, for example, are strongly influenced by the composition of the parish population served by the church. This is reflected in empirical findings indicating that the activities of predominantly black Catholic churches differ from predominantly white Catholic churches [9].

Second, theological innovations and variation that generate sectarian movements do not always result in the creation of new denominations or splinter groups. In the case of Catholics, who account for over one-third of U.S. religious adherents, the development of religious orders provided an outlet for sect-like activity that was contained within the Catholic Church [10]. More importantly, Protestant congregations that can adapt traditional teachings to apply to local conditions are more capable of maintaining vitality among church members and minimizing the possibility of congregational splits [11]. These findings suggest that theological variation exists among congregations within a denomination.

Finally, in post-war America, adherents who shift theology find less need to switch denomination and may simply join a different congregation within the same denomination [12,13]. Recent research has further questioned the notion of socioeconomic, demographic, and geographic homogeneity within denominations [14]. This line of study argues that the greatest degree of population heterogeneity occurs between congregations rather than within them. Congregations, regardless of denomination, possess a greater level of homophily along a variety of dimensions. For example, Reimer [15] finds that social class is a significantly stronger predictor of congregational membership than is denominational affiliation. Another indicator of increasing denominational heterogeneity is seen in recent statistics indicating that 15 percent of Southern Baptist Convention congregations are majority nonwhite congregations [16]. This portends a striking transformation of a once homogenously white religious denomination.

### **3. Bringing the Congregation Back in: Conceptualizing Congregations as an Organizational Population**

Though denominations do influence certain practices in constituent congregations, churches have the capacity to act as autonomous organizations [17]. Even so, only two studies have utilized congregations to measure the religious environment and its link to religious participation. Examining intriguing historical data on rural congregations collected by Brunner [18], Finke and Stark [6] note that the number of congregations per 1000 persons is associated with higher community levels of religious participation. Welch [19] finds that more than one Catholic congregation in the local community influences religious participation among Catholic members. Building on these seminal studies, we contend that the organizational ecology tradition can be drawn on to further explicate the relationship between the religious environment and participation by adherents.

Indeed, others have made similar theoretical connections. For example, Stark and Finke [5] employ the concept of the niche to explain why some denominations have grown in the U.S. and others decline. One of the first adaptations of this perspective to religious congregations linked the concept of niche width to understanding the worship practices of congregations [17]. In a similar vein, Scheitle [20] and Scheitle and Dougherty [21] examine the relationship between niche competition and congregational population dynamics, such as congregational foundings and net change in the size of a denomination. Others have applied organizational ecology to explain historical fluctuations in membership size for specific denominations [22].

Although prior studies have focused on the creation of niche religious markets, researchers have yet to incorporate the concepts of *organizational density* and *density dependence* to explain the working of religious economies. Organizational density refers to the size of an organizational population in a given environment. For us, religious congregations comprise the organizational population of a religious economy. Organizational ecologists posit that the dynamics of an organizational population are dependent on the density of the population [23]. Density dependence occurs because the level of density in an organizational population determines the level of competition between organizations. In turn, competition reduces the rate of organizational founding (the creation of a new organization) and increases the mortality rate of organizations.

Density dependent processes are important because they have important consequences for the composition of an organizational population. Barron [24] notes that organizational density has a direct effect on the average size of organizations. As an organizational population reaches a peak level of density, high levels of organizational founding lower the average organizational size due to the large number of small organizations. As organizational density declines from its peak, average organizational size increases due to the "liability of smallness" [25]. The "liability of smallness" refers to the higher rate of mortality experienced by smaller organizations in dense, highly competitive organizational populations. Thus, low levels of organizational density result in a propensity of large organizations. In contrast, high levels of density result in a large number of small organizations and a smaller average organizational size.

### **4. Institutional Effects of Congregational Density on Religious Participation**

We draw on the notion of organizational density to define our concept of congregational density. Congregational density is defined as the number of local congregations relative to the local population. Our application of organizational ecology posits that a high level of congregational density will result in smaller average congregation size due to density dependent processes. The concept of congregational density is especially useful for framing the individual's religious participation as embedded in the local population of congregations in a religious environment. Density impacts participation by conditioning two aspects of congregational life: participatory structures and social networks.

### *4.1. Participatory Structures*

An important result of smaller organizational size is that congregations develop fewer authority hierarchies and a smaller division of labor [26]. This enables members of smaller organizations to participate more directly than members of larger organizations. In the small congregational setting, this means that members may participate more in decision-making and problem-solving activities. In large congregations, the absence of direct participatory involvement occurs in part because of formalized roles and a greater division of labor among members. Hierarchies are created to manage the large number of activities of the larger congregation. This may result in the establishment of leadership positions filled by additional clergy or lay members, increasing the social distance between rank-and-file members and the congregational leadership. Efforts to minimize this distance using small groups do not eliminate the negative effect of large size on participation and social support [27,28].

Variation in participatory structures has important implications for religious participation. Research on voluntary associations has demonstrated that organizational commitment is directly associated with three aspects of an organization's authority structure: participation of members in decision-making activity, frequency of communication between leadership and members, and the distribution of power within the organization [29]. In turn, the capacity for members to participate in decision-making strengthens the effectiveness of a voluntary organization [30]. Active church participation in terms of service attendance, membership in church groups, and the degree of communication with other members is also correlated with the amount of control and the distribution of control in the congregation [31]. In sum, congregational density results in religious participation through greater organizational commitment created by more participatory structures.

### *4.2. Social Networks*

A second mechanism through which size influences participation is social networks. A number of writers have utilized social network explanations to link the theological orientation of the congregational population to a variety of social outcomes [32–35]. Yet, the impact of congregational density on social networks is less developed. Theories of social organization posit that increasing organizational size erodes group consensus on norms, reduces communication among members, and increases deviance from group norms [26]. In small organizations, strong shared consensus among members provides an accountability structure to regulate members. In addition to social regulation, small size limits the organization's capacity to accommodate differentiation among members [36]. The lack of differentiation results in a homophilous membership base where members share social ties with similar members [37]. Shared norms and homogeneity within the organizational membership also facilitate the development of bonding social capital that may isolate members from the broader community and generate network closure among members [38].

In the congregational context, the homogenous nature of small congregations creates something of a boundary between the congregation and the broader community. As a consequence, congregations foster a high level of network closure, providing a means to develop trust and accountability [39]. One important institutional effect of congregational density is the local primacy given to religion. Network closure makes religion a more vital and central institution in the community because social networks will be disproportionately based on intracongregational ties. A second effect of network closure is the capacity for social control. No matter the norms of denominations, congregations hold expectations for members' participation in church activities [40]. When members are embedded in social networks with dense ties to other congregation members, levels of participation may increase because members mutually reinforce norms of participation [41]. In larger congregations, the capacity for social control of members is weakened because interactions between members become more impersonal.

### **5. Hypothesis: Congregational Density and Religious Participation**

Rather than focus on the distribution of adherents across denominations, we apply organizational ecology theory to understand the local implications of the size of the congregational population. We propose a *congregational density thesis* and hypothesize that the number of congregations relative to the residential population of a community encourages religious participation. This hypothesis is distinct from denominational approaches because it focuses on the institutional effects of religious organizations. Organizational density exerts a downward pressure on the size of congregations due to the density dependence. In turn, congregationally dense religious environments with an abundance of small congregations take advantage of two institutional mechanisms: participatory structures and network closure.

Congregational density enhances participation through horizontal authority structures of smaller congregations. A lack of bureaucratic structure and more direct channels of communication between leaders and members results in higher levels of participation. Participatory structures enhance commitment to the congregation, and members are more likely to become religiously engaged. In addition, network closure increases religious participation because the capacity for social control and accountability among members is enhanced. The greater focus of network ties on religious congregations promotes the salience of religion in community life and underscores the importance of identifying with a congregation for community residents.

### **6. Data and Methodological Section**

To evaluate our hypotheses, we analyze data from the 2000 Social Capital Community Benchmark Survey (SCCBS) merged with religious environment measures from the 1990 and 2000 Religious Congregations and Membership Study (RCMS). The RCMS provides county-level counts of the number of congregations and adherents for 149 religious denominations and religious bodies [42] 1 . The SCCBS data were collected in 2000–2001 using random digit dialing telephone interviews and devised by the Saguaro Seminar at the John F. Kennedy School of Government at Harvard University [43]. Prior studies using these data have focused on topics including social capital, volunteering, financial giving, and public health [44–48].

We link county-level information on the congregational population and contextual population data to the communities of individual respondents. Our analytic sample is limited to 20,723 respondents 2 nested in 258 communities 3 . All analyses use person-level sampling weights.

### *6.1. Dependent Variables: Religious Participation*

The outcomes in our analysis gauge an individual's religious participation across a number of dimensions. Researchers in the sociology of religion have examined the relationship between religious pluralism and religious adherence at the ecological level [49–52]. Our approach extends this prior work and differs from it by employing a multilevel design to test whether community-level religion variables influence individual-level religious participation. This strategy allows us to evaluate the importance of the community level in relation to the individual level, rather than limiting ourselves to inferring individual-level processes from aggregate measures. Only one study to our knowledge has hierarchically linked community-level data on religion to data on individual religious participation. Borgonovi [44] examines the relationship between denominational pluralism and religious participation. In that study, religious participation was measured using three indicators: weekly attendance at religious services, religious volunteering, and religious giving.

We build on this measurement strategy by incorporating six measures of participation: (1) volunteering for a religious charity (26%), (2) membership in a non-church religious organization (15%), (3) church membership (55%), (4) participation in church activities such as choir, prayer meetings, and bible study (38%), (5) weekly worship service attendance (41%), and 6) charitable giving to church or religious causes (67%). Our focal dependent variable is a summary index (α = 0.84) of the standardized scores on each of the six indicators [43]. We also perform separate analyses for binary measures of each individual item to assess the strength of our findings.

<sup>1</sup> An approximate census of U.S. religion, the RCMS data provide the most thorough record available of religious adherents and congregations by counties. The RCMS was sponsored by the Association of Statisticians of Religious Bodies in America (ASARB).

<sup>2</sup> Missing values on the individual-level measure of household income were imputed using conditional mean imputation.

<sup>3</sup> Rural South Dakota is excluded because geographic identifiers are not present for these respondents.

### *6.2. Contextual-Level Independent Variables*

The independent variables in our analysis include characteristics of individual survey respondents and community-level variables. All measures are from the year 2000 unless otherwise indicated. The key community-level independent variables in our analysis are measures of the religious environment. These are derived from the 1990 and 2000 RCMS in the U.S. data [42]. We calculate three measures from county-level tabulations of religious congregations and adherents: congregational density, denominational pluralism, and the percent of population that is a church member.

We measure congregational density as the number of religious congregations per 1000 county residents. Our exploratory analyses identified skewness in the distribution of congregational density. Therefore, a natural log transformation was performed on this variable to induce normality. Our denominational pluralism measure is based on the Hirschman-Herfindahl Index of concentration. Denominational pluralism is calculated as:

$$DP = 1 - \sum\_{i=1}^{}^{2} p\_i^2 \tag{1}$$

where *pi* is the proportion of all religious adherents that belong to a given denomination. Large values of the denominational pluralism index indicate that religious adherents are dispersed across a number of denominations and low values signal that a specific denomination has a large market share of adherents.

We account for potential endogeneity in our models by controlling for community-level church membership in 1990. This variable comes from the 1990 version of the RCMS and is calculated as the number of religious adherents in the county divided by the total community population. We include this measure to address the potential reverse causation in our models due to the possibility that high levels of religious involvement among community members may result in a large number of congregations per person.

To control for socioeconomic and demographic characteristics of the county, we combine several items from the 2000 Census of Population and Housing and the Office of Management and Budget. We construct a summary measure that captures socioeconomic diversity in the composition of the county population. It is based on the Gini coefficient of family income inequality, the percentage of the population aged 25 and older with a high school diploma, and the percentage in poverty. Due to high correlations among these four variables, we summarize them using a factor score we refer to as a Socioeconomic Disadvantage Index (α = 0.87). We control for socioeconomic disadvantage because individuals' religiosity may be higher in contexts experiencing large socioeconomic disparities and insecurities [53,54]. Disadvantage should also positively relate to religious charity activities because in a highly disadvantaged community, congregations may be more involved in social welfare activities [55].

**23** 

To be sure, we account for major theoretical concepts at the community level intentionally using only a few variables. However, this parsimonious specification of models reduces the potential for multicollinearity at the community level 4 .

### *6.3. Individual-Level Control Variables*

Our individual-level control variables capture demographic, socioeconomic, and religious characteristics associated with religious participation in the literature [55–59]. To account for demographic variability in religious participation, we include a continuous measure of age in years and binary indicators for the respondent's gender, race-ethnicity, marital status, and the presence of children under the age of six in the household. Our measure of race-ethnicity includes categories for persons of Hispanic origin, Nonhispanic White (reference), Nonhispanic Black, and Nonhispanic persons of other race groups. At the individual level, the socioeconomic control variables in our models include measures of education, household income, and labor force status. Education is measured using four categories: less than high school diploma (reference), high school graduate, some college-associate degree, and college graduate. Labor force status is classified as employed, unemployed, and not in the labor force (reference). We classify religious affiliation using a coding scheme based on the Steensland *et al.* [60] typology of religious adherents (*i.e.*, Conservative Protestant, Mainline Protestant, Catholic, and Other). Predominantly black denominations are classified as Mainline Protestant due to the small number of persons identifying affiliation with this group. We pooled these categories because our models control for race-ethnicity which is strongly correlated with membership in a predominantly black denomination. To account for geographic differences in religious participation we control for southern residence (South = 1). Researchers have noted regional variation in church attendance, especially between the South and other regions [61–63]. Summary statistics are presented in Table 1.


**Table 1.** Descriptive statistics.

4 No evidence of multicollinearity was found upon examination of a correlation matrix with our community-level independent variables. All bivariate Pearson correlations are less than 0.50 (see Appendix). Variance Inflation Factors are below 1.5.


**Table 1.** *Cont.*

Note: Individual-level data are weighted.

### *6.4. Analytical Strategy*

Due to the nested structure of the data, we use multilevel modeling to test our hypotheses regarding the effect of congregational density on religious participation 5 . Multilevel modeling is ideal for our hypotheses and data because the technique takes into account dependence of individual-level cases nested within the same contextual-level unit and properly estimates standard errors and cross-level relationships in hierarchical data [64]. Models are fitted using HLM 6. Our first model examines the composite religious participation measure. This model includes our lagged measure of church membership from 1990. In addition to providing a robust test of our hypotheses, this model assesses the degree to which reverse causation may impact our findings. We then estimate separate hierarchical logistic regressions for each of the six components of the

<sup>5</sup> Prior to estimating multilevel models with predictors, we examined an unconditional model to confirm the presence of significant variation in religious participation across communities (ICC = 0.0436; χ<sup>2</sup> = 1538.52481, *P* < 0.001).

religious participation index (religious volunteering, non-church religious organizations, church membership, church activities participation, weekly service attendance, and religious giving). These supplementary models demonstrate the degree to which our hypothesized congregational density effect occurs across a diverse set of religious participation indicators.

### **7. Results**

In Table 2, we report regression coefficients predicting the composite religious participation measure. This model tests the notion that congregational density (the number of congregations per 1000 persons) is positively associated with religious participation. The results are consistent with our expectations. At the community level, an increase in the number of churches per 1000 persons is significantly associated with a higher score on the composite religious participation measure (b = 0.083; *P* < 0.001). By including a time-lagged measure of religious membership, the percentage of the population belonging to a church in 1990, this model also assesses the possibility that the relationship between congregational density and religious participation is due to reverse causation, such that more religiously involved communities contain more churches per 1000 persons. We find that the coefficient for congregational density remains positive and significant net of lagged percent church members. This finding suggests that the effect of congregational density is robust and cannot be attributed to the level of religiosity within a community.


**Table 2.** Multilevel model predicting composite religious participation index.


**Table 2.** *Cont.* 

Note: Community *N =* 258; Individual *N =* 20,723; Individual-level data are weighted; \* *P* < 0.05; \*\* *P* < 0.01; \*\*\* *P* < 0.001 (two-tailed tests).

To assess the relative strength of significant relationships, we use the standard deviations of county-level means to calculate standardized coefficients. Of our two focal county-level indicators, the magnitude of the congregational density effect is clearly the largest: the standardized effect of congregational density is 0.335 (0.083 × 0.635/√ 0.0247) and denominational pluralism is 0.303 (0.003 × 15.797/√ 0.0247).

Other significant community-level effects are also present. In this model, we also test the denominational pluralism hypothesis. This hypothesis suggests that increases in denominational pluralism are associated with higher levels of religious participation. The denominational pluralism index is positively and significantly associated with religious participation. This relationship provides evidence that persons residing in denominationally heterogeneous communities are more highly religiously involved. Inconsistent with insecurity theorists [53], socioeconomic disadvantage does not have a significant positive association with religious participation. Religious participation is not a function of community-level disadvantage and insecurity in these data. This finding is interesting because it demonstrates that the effect of socioeconomic status is limited to the individual level, where, as we will see below, increased individual-level socioeconomic status is positively related to religious participation.

This model also includes our set of individual-level control variables and we observe a number of statistically significant effects. We find that older individuals, females, married persons, and those residing in households with a child under the age of six exhibit higher levels of religious involvement. The effects of these demographic measures on religious participation are consistent with those of Alston and McIntosh [65] and Cornwall [66]. In contrast to Alston *et al*. [65] and Hoge *et al*. [67] who find little effect of socioeconomic status on religious participation, our results indicate that employment status is negatively related to religiosity, whereas household income and education have positive relationships with religious involvement. The coefficients for educational attainment indicate that persons with post high school education are more likely to participate in religious activities than high school graduates and those not graduating from high school. In terms of race and ethnicity, our findings echo those of Ellison and Sherkat [61] who report higher rates of religious participation among Nonhispanic African Americans.

In Table 3, we further evaluate the congregational density thesis by fitting hierarchical Bernoulli logit models for each indicator comprising our composite religious participation index. For brevity, only the coefficients for the congregational density, denominational pluralism, and the lagged church membership variables are presented. The estimates for these variables are largely consistent with those in Table 2. The first thing to note in Table 3 is that congregational density exhibits a consistent effect across all six dimensions of religious participation, even when holding other factors constant. Illustratively, compared with residents of the most congregationally sparse context, people in the most congregationally dense community are 2.4 times (*e* 1.935 × 0.296*/e* <sup>−</sup>0.979 × 0.296) as likely to volunteer, 1.6 times as apt to belong to a religious organization, 2.3 times as likely to be a church member, almost twice as prone to participate in church activities, one and a half times as likely to attend weekly, and 1.6 times as liable to financially give to religious charities. We also find that denominational pluralism has a significant positive effect on four of the six outcomes. Pluralism does not influence church membership or religious giving. This latter set of results dovetails with Borgonovi's [44] recent observation of a positive relationship between county-level religious pluralisms and religious volunteering, but is not consistent with her observation that pluralism is significantly associated with giving but not service attendance. The lagged percent church members control variable is significantly and positively related to all six dependent variables.



NOTE: Models also include community-level socioeconomic disadvantage index. Individual-level data are weighted and measures controlled are: age, sex, marital status, presence of children under age 6, labor force status, household income, educational attainment, race/ethnicity, denominational affiliation, and southern residence; \* *P* < 0.05; \*\* *P* < 0.01; \*\*\* *P* < 0.001 (two-tailed tests)

Calculating the magnitude of our focal county-level variables again reveals congregational density to have a stronger effect than denominational pluralism. A one-standard-deviation increase in congregational density is associated with a 20.7% (*e* 0.007 × 0.15.797) boost in the odds of religious volunteering, whereas a one-standard-deviation increase in pluralism is associated with a 13.7% (*e* 0.008 × 15.797) increment in the odds of volunteering—a substantial difference in odds (7%). Congregational density has a standardized effect of an 11.2% increase in the odds of non-church religious organization membership and pluralism 10.8%. For participation in church activities other than worship service, the standardized effects are 13.3% (congregational density) and 12% (denominational pluralism). The magnitude of the effect of congregational density on individual-level church membership is also considerable (20%). However, pluralism is not significant for church membership; people are just as likely to be a member of a congregation in a denominationally heterogeneous community as a community with a relatively high level of religious monopoly. In the case of weekly church attendance, pluralism has the stronger effect: a one-standard-deviation increase in pluralism is associated with a 12.4% and density a 9.2% increase in the odds of weekly church-going. As density goes up by one standard deviation, religious giving's odds also rise by roughly 11%. Pluralism has no significant influence on giving. Over and above the influence of community-level factors such as religious pluralism and individual-level characteristics, not only does congregational density play a significant role in all of the religious behavior outcomes, but with the exception of one indicator, congregational density has the strongest effect of our two focal community-level predictors.

### **8. Conclusions**

Two lines of argument in the sociology of religion tradition have suggested that the distribution of adherents across denominations affects local levels of religious involvement [2,4,68]. Implicit in the use of denominations is the assumption that denominations act as religious establishments themselves and adapt to fit the needs of population segments in a community. Instead, we suggest that, regardless of the local denominational composition, the number of congregations within a community will largely determine the level of competition.

The empirical evidence presented here for a congregational density perspective is reasonably compelling in three key ways. First, after accounting for a variety of individual and contextual-level covariates, congregational density remains a significant predictor of religious participation. This relationship holds after controls are introduced for demographic, socioeconomic, and spatial factors. Second, individual-level covariates have a substantial impact on the level of religious participation. This is a critical finding, because with one exception [44], all tests of the religious economy argument have utilized aggregate data. Our findings suggest that the analysis of aggregated rates of participation that do not adjust for individual-level attributes may be incomplete. Third, our findings regarding congregational density are obtained after accounting for the effect of denominational pluralism. Unlike recent research that suggests no relationship between religious pluralism and religious participation [44], we find support for the religious economies perspective forwarded by Finke and Stark [6]. Our models also indicate that congregational density has an effect independent of denominational pluralism and suggests that our congregational density argument makes a unique contribution to explanations of religious participation 6 .

<sup>6</sup> We note that the communities included in the Social Capital Community Benchmark Survey were not randomly selected. Thus, the sample may not reflect the entire national population.

An important contribution of the congregational density approach is that it bridges the denominationally based approach of religious economy scholars and the emerging line of study on congregations [69]. The congregational density argument derived from organizational ecology theory is largely compatible with many of religious economy propositions on group dynamics and competition (for a complete list, see Appendix of Stark and Finke [5]). To be sure, Stark, Finke, and colleagues note the role of congregational size and recognize the consequences of social organization on member commitment.

Despite these similarities, our approach differs on three key aspects of religious economy theory. First, our congregational density perspective does not rest on a sect-church differentiation in which congregations can be placed along a continuum of tension with the broader society. We do not condition religious commitment and participation on the degree of tension between the congregation and the community. Our argument focuses on the relationship between the size of the local congregational population and the social organization of congregations. The results here show that congregational density exerts a significant influence on religious participation that is independent of denominational pluralism or levels of religiousness from prior decades. The congregational density approach differs with religious economy theory in a second way by emphasizing the importance of the organizational environment. The key environmental factor for religious economy theory is the level of regulation exerted on the religious marketplace by political forces. Beyond state regulation, there is limited discussion of features of the organizational environment that might explain the level of religious participation in a given community. Our density based approach begins to address this issue. Finally, we argue that the assumptions underlying our congregational density approach are not influenced by ongoing changes in U.S. denominational affiliation. Although survey data on religious affiliation point to a decline over the past few decades, the number of congregations in the U.S. between 1980 and 2000 grew at a rate of 14.2% as compared to that of the U.S. population (13.2%) during this time (ARDA).

The results presented in this paper also point to an important avenue for future research. Although researchers have used denominational measures of the religious environment, such as the percentage of the population adhering to a conservative protestant religious tradition (for examples, see Beyerlein and Hipp [32]; Ellison, Burr and McCall [70]), much less is known about the relationship between congregations and measures of community welfare. The concept of congregational density is promising because it provides an institutional explanation of how the religious environment influences nonreligious sectors of community life. It also raises an important question. Given that an organizationally dense congregational environment fosters the development of participatory structures, is the religious environment associated with the local civic culture and democratic participation in community governance? We suspect that the social organization of the religious sector covaries with a participatory environment within nonreligious voluntary associations and other community problem-solving structures. Thus, congregationally dense communities may be better equipped to address the needs of local residents in terms of a broad array of health, safety, and socioeconomic well-being factors. These may be promising topics for further inquiry.

### **Author Contributions**

The authors jointly conceptualized, analyzed data, and produced this manuscript.

### **Appendix**


**Table A1.** Bivariate correlations for community-level variables.

Note: \* *P* < 0.05; \*\* *P* < 0.01; \*\*\* *P* < 0.001 (two-tailed tests).

### **Conflicts of Interest**

The authors declare no conflict of interest.

### **References**


### **Go Forth and Multiply: Revisiting Religion and Fertility in the United States, 1984-2008 †**

### **Casey Borch, Matthew West and Gordon Gauchat**

**Abstract:** Many studies on the fertility differential by religion have considered both Catholics and Protestants to be equally homogenous groups. Contrary to these studies, we contend that Protestant fertility must be studied in the context of heterogeneous groups. Specifically, conservative Protestantism, with its beliefs about artificial birth control mirroring Catholic teaching, should be examined separately from other Protestant traditions. Using data from the General Social Survey we find that conservative Protestants and Catholics had about the same level of fertility, while mainline Protestants have a fertility rate that is significantly lower than that of Catholics. We also examine the changes in these differences over time.

Reprinted from *Religions*. Cite as: Borch, C.; West, M.; Gauchat, G. Go Forth and Multiply: Revisiting Religion and Fertility in the United States, 1984-2008 † . *Religions* **2011**, *2*, 469–484.

### **1. Introduction**

Membership in a religious tradition or denomination can affect fertility [1,2]. Historically, much research has aimed to describe the intersection of religion and fertility in the United States [3-7]. This research has focused on the differential fertility of Protestants and Catholics, but has largely ignored the difference between Protestant sub-groups. In particular, there has been very little research looking at the effect of being a conservative Protestant on fertility compared with other religious groups. Following the work of Woodberry and Smith [8], we use the term "conservative Protestant" in place of Steensland *et al.*'s [9] "evangelical Protestant." Since "conservative" Protestant includes evangelicals as well as fundamentalists and Pentecostals, it is a more inclusive categorization of the Protestant sub-groups of interest to us.

Conservative Protestantism represents the largest religious tradition in the United States, which corresponds to roughly 30% of the U.S. population [8,10]. Hout and Fischer [11] found that conservative Christians tend to have higher fertility than non-conservatives, and thus predicted a growth of conservative religious traditions over time, which includes conservative Protestants. One prominent study looking at the fertility differences between U.S. Protestants concluded that evangelical Protestants will be the next promising arena for the study of American fertility behavior ([6], p. 542):

 Recent surveys have found an increase in religious interest in the United States, probably due in part to the so-called evangelical movement. Yet almost nothing is known about the fertility consequences of this phenomenon. As Catholic and Protestant fertility levels continue to converge (Westoff and Jones), this and other dimensions of Protestant life seem promising choices as new arenas for the study of American fertility behavior.

The primary purpose of this study is to examine the association between conservatism and fertility in the United States. Are the higher levels of fertility found among conservative Protestants due to intrinsic components of their religious tradition, or are they a statistical artifact created by an underlying demographic structure? That is, does religious affiliation explain their high levels of fertility, or does the characteristic hypothesis—which asserts underlying demographic characteristics are the primary predictors of fertility—provide more accurate predictors of their fertility?

By examining this association, we will add to the literature in two important ways. First, currently there is little research applying multivariate models to explain the relationship between conservative Protestants and fertility behavior. This study hopes to re-open research on religion and fertility in the U.S. specifically focusing on conservative Protestants. Second, this study will explore future fertility trends of conservative Protestants, which will have implications relating to the future of the U.S. religious and political landscape. Future fertility trends are important because of conservatives' influence as a religious and political movement, and their growth may affect their efficacy as a social movement [12].

### **2. Background**

Religious affiliation and fertility were formerly at the forefront of demographic research in the United States. Pearce [13] posits that, given the many studies demonstrating that the number of children in a family varies from religion to religion, it is reasonable to assume that there is a connection between religion and fertility. Most research in this tradition emphasizes the discrepant fertility patterns between Catholics and Protestants [14,15]. These studies, along with those focusing on the Jewish population (e.g., [16]), found Catholicism to be more highly correlated with high fertility rates compared to Protestants, Jews, and the non-religious. One of the strongest suggested reasons for high fertility among Catholics was the Church's teaching on contraception and abortion, which strongly prohibited artificial birth control (excluding the rhythm method). Westoff and Jones [7] found that Catholic and non-Catholic fertility rates in America were slowly converging until the baby boom occurred. During the 1950s and early 1960s, Catholic fertility skyrocketed compared to other religions, significantly widening the gap not previously seen in the 19th century. By the time Vatican II occurred and the papal encyclical on birth control became public, about one-third of American Catholic women were on the birth control pill [14]. Lenski [17], among others (e.g., [18]), found that the extent to which church members are willing to reform their birth control practices changed the way Catholic leaders thought about "authority" and "dissent." It is at this time when the Catholic Church began to intervene in the sexual behaviors of the married faithful. This led parishioners to begin to disobey church teaching on contraception, among other things.

The final outcome of these changes was Catholic fertility rates matching that of other religious denominations. Westoff and Jones ([7], p. 209) pinnacled this phenomenon with a paper entitled "The End of 'Catholic' Fertility," concluding that, although Catholics in the mid 20th century had slightly higher fertility than their non-Catholic counterparts at this time, towards the mid-1970s "the two trends nearly come together," effectively ending a uniquely Catholic fertility rate. With the apparent end of a "Catholic" fertility having thus occurred, interest in religious affiliation and

**37** 

how it affects fertility withered away as discrepancies in fertility levels between religious groups approached zero.

Along these lines, work by Bartkowski, Xu, and Levin [19] and more specifically Mahoney *et al.* [20] suggests that "sanctification theory" helps to explain the convergence. Mahoney ([20], p. 222-223) notes that believers "view family relationships as sacred… Judeo-Christian religions portray the burdens and pleasures of parenting as opportunities to model and deepen one's understanding of God's love, patience, and commitment, and frame the parental role as a sacred calling that requires personal sacrifices." Both the Catholic and conservative Protestant faiths have strong emphases on child-rearing as a vocation, responsibility, and calling from God to do their duty and treat children as a gift. By having more children, then, it may be perceived that a couple is being a "good Christian." This is particularly true for conservative Protestants because of their strong belief in the Bible as the literal word of God and the various Biblical commands to "multiply," populate the earth, and the like (as well as God's anger at those who attempt birth control, such as Onan's withdrawal). Thus, more conservative denominations likely place more emphasis on the Bible and its pronatalist worldview. Scripture, along with sin and salvation, constitute Bartkowski's "three S's" that differentiate conservative Protestants from all others [21]. The authors note that while sanctification of the family may be good in that it promotes social cohesion, it can also prove negative in situations such as the family whose parents are of differing religions.

Past research comparing Protestant and Catholic fertility tended to lump all Protestants together much in the same way as Catholics are lumped together—that is, as one unitary Protestant Church similar to the one Catholic Church—and thus did not pick up the higher levels of fertility occurring for conservative Protestants [3,4,7]. This collapsing of Protestant denominations into one conceptual group risks the validity of studies comparing fertility rates by religion. For example, Bean *et al.* ([22], p. 91) lamented how past studies typically classify respondents as "Protestants, Catholics, and Jews; these categories are not homogenous enough to provide a fair test of the relation between religion and fertility" which was particularly problematic with their sample of Utah Mormons. In addition, Pearce ([13], p. 20) notes that "within Protestantism, studies show substantial variation in fertility across affiliations." Possible explanations for such variation include demographic transition theory and the minority status hypothesis (for a comprehensive review, see [13]).

When conservative Protestants are distinguished from the rest of mainline Protestants, their fertility consistently remains higher than that of mainline Protestants and Catholics. Figure 1 shows the average number of reported children for conservatives, mainline Protestants, and Catholics. All three denominations witnessed a drop in the average number of reported children over the last three decades. This result is consistent with prior research that finds all major religious communities in the United States witnessed a drop in the expected family size as the second demographic revolution occurred [23,24]. Conservatives have largely maintained a higher average number of reported children since 1984 than Catholics and Mainline Protestants. Indeed, this mirrors Hout and Fischer's [11] work that finds that the vast majority of conservative Christian growth is due to higher fertility and not conversion. As Greeley [25] hypothesized in his monograph *Religion in the Year 2000*, growth of a particular religion or denomination will most certainly be determined by the number of present-day adherents' children, not by conversion rates. If conservative Protestants have large families with many children, while Mainline Protestants have smaller families, then conservatives should theoretically have the advantage of numbers in the proceeding generation. This is especially true for conservatives who begin parenting at markedly younger ages than mainline denominations, thus allowing more time for family growth early in the life course [26].

### *2.1. Who are the Conservative Protestants?*

Conservative Protestantism is flourishing in the United States while liberal Protestant church membership is in decline [27]. The literature tells that conservative Protestantism acquires and maintains members because they demand complete loyalty, unwavering belief, and rigid adherence to a specified way of life [28,29]. Conservative Protestantism is said to have developed in the 1970s as a reaction to the social revolution which occurred in the previous decade. Conservative Protestants believe that the greater personal freedoms in gender roles and sexuality which became prevalent in the previous decade were undermining their core values [30]. Also, Christian Right activism and evangelical growth focused on resistance to moral relativism in the surrounding culture. Indeed, as Emerson and Hartman ([31], p. 127) write: "Without modernization and secularization [of this century] there would be no fundamentalism…" Membership in evangelical groups has risen significantly since the initial push for activism initiated in the 1970s, with conservative Protestants making up more than a "quarter of the American population" in 1998 and varying from many Americans in their unique opinions on "…gender-roles, childrearing styles, [and] political orientation…" ([8], p. 25). Shibley [32] outlines three components for defining contemporary conservative Protestantism: (1) Have had a born-again experience resulting in a personal relationship with Jesus Christ, (2) Accept the full authority of the bible in matters of faith and in daily conduct, and (3) Are committed to spreading the gospel by bearing public witness to their faith. These three components provide the common thread for all conservative Protestants.

### *2.2. Conservative Protestants and Fertility*

Theory on religious influences on fertility coupled with prior research on conservative Protestants suggests that there are good reasons to believe that the religious beliefs and practices of conservatives influence their fertility. Current theory on religion and fertility states that there are three components that allow religion to influence fertility behavior [1]. First, religions articulate behavioral norms that have linkages to fertility outcomes (e.g., directive on contraception or teachings related to gender roles and family life). Second, religious groups possess the means to communicate its teachings to its members and enforce compliance. Lastly, members with a strong sense of attachment to the religious community will likely adhere with its norms and teachings. By comparing previous literature studying conservative Protestants with the aforementioned theory on fertility and religion, a strong case can be made to infer that conservative Protestantism influences fertility.

The first component is present with contemporary conservative Protestant churches advocating strong prohibitive norms concerning sexuality, gender roles, and family values [8,32]. Also, conservatives look to the Bible for matters of faith and daily conduct, in which there is considerable evidence that the Bible advocates having numerous children. For example, Genesis 1:28 states "Have many children so that your descendants will live all over the earth and bring it under their control." The second component highlights communication and enforcement of general norms. Again previous research has found conservative Protestants largely possess these means as they acquire and maintain members by demanding loyalty, unwavering belief, and rigid adherence to a specified way of life [28,29]. The last component requires a strong sense of attachment to the religious community, which conservative Protestants possess as evident by the popularity of their new market oriented places of worship [32]. In addition, evangelicalism is a religious movement created by feelings of alienation from what they view as an increasing secularized world. These feelings of alienation coupled with numerous popular congregations, in which like-minded individuals attend, provide compelling evidence that many conservative Protestants will feel a strong sense of attachment to their religious community. Overall, there is reasonable evidence to suggest that conservative Protestantism will influence fertility.

### **3. Data and Methods**

### *3.1. Data*

The data come from the General Social Survey (GSS). The GSS is based on a probability sample of the adult civilian population of the United States, stratified by region and metropolitan *versus* non-metropolitan residence. The GSS continues to use in-person interviews, and sampling techniques and response rates are well documented [33]. A major advantage of the GSS is that most of the questions keep the same wording from year to year, so data can easily be compared longitudinally or pooled across years. As per the suggestions in the GSS technical information, sample weights were not used in our analyses. This is because from 1984-2002 the data were taken from a full probability sample of households in the U.S., thus making the GSS self-weighting. Also, as suggested, we did not include data from 1987, because of an over-sample of Blacks. After 2004, weights are necessary only if the data are analyzed within year. Thus, since our focus is longitudinal and we include dummy variables for each year in our models, we do not have to include sample weights in our models.

### *3.2. Variables*

The dependent variable is self-reported fertility. The GSS asked respondents the following question [GSS mnemonic = CHILDS]: "How many children have you ever had? Please count all that were born alive at any time (including any you had from a previous marriage)." The response categories ranged from zero to eight or more children. This question was asked in 17 years over the period 1984 to 2008.

The following independent variables were included in the models. We measured sex as a dummy variable for *female*. Although the original dataset included white and an "other" category, we measured race as a dummy variable distinguishing *black* from all other races. *Age* was measured continuously by the GSS with a range of 18 to 89 years old. To account for expected non-linearity in the effect of age on fertility, we also included *age-squared* in the models. *Education* was included as a continuous variable that ranged from 0 to 20 years of schooling. Marital status is a dichotomous variable that distinguished those *married* were from those unmarried. We treated household *income* (in real 2000 dollars) as a continuous variable. A dummy variable to distinguish respondents living in the *South* from those living elsewhere was created and included in the models. The specific delineation of the states into these regions was determined by the way in which the GSS partitioned the states into regional categories (see Davis *et al.* [33] for further explanation). We also compared those living in *rural* areas from those living elsewhere (non-rural). As differentiated by Tuch [34,35], rural included the following areas: Not within an SMSA or standard metropolitan statistical area, (within a county) and—a small city (10,000 to 49,999); a town or village (2,500 to 9,999); an incorporated area less than 2,500 or an unincorporated area of 1,000 to 2,499; open country within larger civil divisions, e.g., township, division. Finally, we use Steensland *et al.*'s [9] typology to place GSS respondents into religious groups. These groups are as follows: *Catholics*, those with *no religious affiliation*, those claiming religious faiths *other than Christian*. The Protestant group was further separated into *conservative Protestants*, *Black Protestants*, and *other Protestants*. Table 1 presents descriptive statistics for all the variables discussed above along with means by decade and correlations with time.

### *3.3. Analytic Strategy*

Since the dependent variable is a count variable (*i.e.*, number of children), the most appropriate method of analysis is negative binomial or Poisson regression. The model takes the following form: log(m) = a + bx, where the dependent variable (y) and the mean of y (m) are related by the Poisson distribution (error). Unlike ordinary regression, however, the variance of the error term is fixed. In an ordinary regression, the error variance is estimated from the model. With the Poisson distribution, the error variance is equal to m (the mean). If the Poisson distribution is appropriate, the deviance will have a chi-square distribution with N-k degrees of freedom, where N is the number of cases (not the total number of events) and k is the number of parameters fitted. In effect, Poisson regression is a log-linear model because it uses the log link.

The restriction of variance = mean (σ<sup>2</sup> = µ) is often hard to satisfy. In fact, our data are what is known as "over-dispersed." This occurs when the variance is greater than the mean (σ<sup>2</sup> > µ), see Table 1. Running a Poisson model on over-dispersed data can produce unreliable results. If the error is uncorrelated with independent variables, the estimates of the parameters will be unbiased and consistent. However, the standard errors will be underestimated. To avoid this pitfall, methodologists suggest that the negative binomial model is the most appropriate method of analyzing count data that are over-dispersed. The coefficients from negative binomial are interpreted the same as those from Poisson regression: (eB−1)\*100, for percent change, and eB, for estimated counts. Negative binomial assumes that the systematic part is log(m) = a + bx + u, where u is a random variable with a gamma distribution. The relationship between m and y is again given by a Poisson distribution. The combination of these two assumptions gives you a particular distribution, the negative binomial. The assumption of a gamma distribution for u is somewhat arbitrary. Basically, the assumption is made because it gives a reasonably simple and well known distribution when combined with Poisson. In sum, using the negative binomial model will give more efficient estimates when the distribution is over-dispersed.


**Table 1.** Descriptive Statistics for all Variables in the Models, 1984-2008 (N = 36,020).

Note: Standard errors are in parentheses. \* p < 0.05

### **4. Results**

### *4.1. General Trends*

Figure 1 displays annual change in mean reported fertility by religious group. Fertility is simply the number of children reported by respondents; it ranges from zero children to eight or more children. The ceiling of eight children was set by the GSS; however this does help to alleviate some of the skew in the variable (skewedness = 1.107; Kurtosis = 1.174). Reported fertility (hereafter, we will use "fertility" in place of "reported fertility") decreases steadily from the mid-1980s to the late 1990s before rebounding after the turn of the century; particularly among Catholics.

**Figure 1.** Average Reported Number of Children by Year and by Religious Group.

The trends by religious group basically follow the trend of overall fertility (*i.e.*, higher fertility in the early period and lower fertility until late in the period); however, there are two noticeable differences. First, the trend for Catholics is lower than that for Protestants. This realigning of fertility rates by religious group has been documented in other demographic research and thus our results reaffirm that work (e.g., [7]). Secondly, the fertility rates of conservative Protestants are higher than that of both Catholics and other Protestant groups. This trend is especially clear in the early- to mid-stages of our analysis period. In the next section, we explore in more detail these trends.

### *4.2. Regression Models*

Table 2 shows parameter estimates and fit statistics from three different models. The first model includes period effects (dummy variables for each year) and demographic variables with no interactions between them. The estimated effects of the demographic variables are reported in Column 1, while the period effects are omitted to save space.

The results in Column 1 show that married people, females, blacks, and those who live in rural areas, and those who attend religious services regularly all report having more children than their specific counterparts. On the other hand, the more educated, the more affluent, and Southerners report having fewer children than those in the alternative groups. The cohort variable shows the expected pattern of those born in the earlier eras have significantly more children than those from Generation Y (born after 1978). The effects of age were non-linear. Specifically, the results suggest that additional years of age increase fertility (effect of age = 0.088, p < 0.001) at a decreasing rate (effect of age-squared = −0.001, p < 0.001). Hence, the age effects take on the expected inverted U shape.


**Table 2.** Results of Three Negative Binomial Models Predicting Self-Reported Fertility.

Note: \* p < 0.05, \*\* p < 0.01, \*\*\* p < 0.001. Standard errors are in parentheses. Dummy variables for each survey Year were omitted to save space. Coefficients were not transformed by exponentiation for presentation. LR χ<sup>2</sup> = deviancenull - deviancefitted, with the number of parameters for the significance test in parentheses.

Model 2 adds predictors for religious affiliation to the period effects and demographic predictors found in Model 1. Specifically, we added dummy variables for Protestant, those claiming other religious faiths, and those reporting no religious ties. The reference group is Catholics. The inclusion of the new predictors did not significantly change the direction or significance of the estimates noted in Model 1. The second model addresses the difference in fertility between the two largest religious denominations, Catholic and Protestant. As expected, on average and net of other predictors, Protestants reported significantly fewer children than Catholics. In fact, our models suggest that Protestants have about (e.058−1)\*100 = 5.6% fewer children than Catholics. Also, those of other religious faiths and no religion had significantly lower fertility than Catholics—however, these groups have fewer respondents in them.

Model 3 further separates the Protestant denomination into four categories—Mainline Protestants (e.g., Presbyterians, Episcopalians, *etc.*), Black Protestants (e.g., National Baptist Convention, National Missionary Baptist Convention, *etc.*), conservative Protestants (e.g., Church of Christ, Pentecostals, *etc.*), and "other" Protestants (Lutherans, Methodists, *etc.*). Again, Catholic is the reference group. Theory and research predicts that conservative Protestants should have lower fertility than Catholics, but higher fertility than mainline Protestants or any of the other Protestant groups. The results for Model 3 support this assertion. Specifically, on average and net of all other predictors, mainline Protestants have significantly lower fertility than Catholics. To explain further, mainline Protestants have about (e.097−1)\*100 = 9.2% fewer children than Catholics. Importantly, the fertility of conservative, Black, and other Protestants was not significantly different than that of Catholics.

With respect to comparing across the religious groups, our results suggest that membership in the different religious groups had differential effects on fertility. Specifically, an F-test for class variable comparison showed that the coefficients for at least one of the religious groups differed from the others F[6,35896] = 20.58, p < 0.001. Subsequent analyses revealed the differences to be mainly associated with the fertility of mainline Protestants. Over the entire time-period the fertility of mainline Protestants was about 1.85 children, while the fertility of conservative Protestants was significantly higher at about 2.03 children. The fertility of Black Protestants was also significantly higher than that of mainline or conservative Protestants at about 2.38 children. Thus, the results show that the fertility of conservative Protestants was significantly *higher* than that of mainline Protestants, but significantly *lower* than that of Catholics or Black Protestants.

The results presented so far are not unique and coincide with many other studies of religion and fertility. However, one area that is largely understudied is the trends in fertility of each religious group over time. According to previous research, conservative Protestants have higher fertility than other groups and that this is what is driving their increasing numbers [6,36,37]. This suggests that fertility rates by religious group should show conservative Protestant fertility to be moving away from that of mainline and other Protestant sects.

In order to test the hypothesis that in the last few decades the fertility rates of conservative Protestants have moved away from that of mainline Protestants and Catholics, it is necessary to include interactions between the key independent variables representing religious denomination and some function of time. The most general form of interaction would treat year as a class variable. Since this approach would involve estimating a large number of parameters, it would have relatively low power. Consequently, it is desirable to use some simple function of time in the interaction term. Since we are interested in gradual shifts, the most obvious possibility is a linear trend term increasing from 0 in 1984 to 24 in 2008. Interactions involving a trend term imply a steady divergence or convergence of groups over the whole period. In reality, change might take more complex forms—for example, the difference between groups might grow for a period of time, but then remain constant. However, the linear trend model provides a useful analytic starting point.

Estimates from a model including interactions with a linear trend are shown in the first and second columns of Table 3. The main effects are shown in the second column, while the interactions are shown in the third column. Our theoretical argument implies that the interactions between religious denomination and trend will be non-zero; that is, the fertility of conservative Protestants is growing relative to Catholics. In contrast to this expectation, the interaction with conservative Protestant is negative and not significant (B = −0.003, p > 0.05). A non-significant coefficient means that the trends in fertility rates for Catholics and conservative Protestants are moving in parallel. Said another way, over time, the gap between Catholics and conservative Protestants has remained the same. However, the gap between Catholics and those with other religious affiliations and those with no religious affiliation is growing larger (B = −0.007, p < 0.05 and B = −0.005, p < 0.05, respectively). This means that, the negative effects of no religion and other religion on fertility are growing stronger over time; said another way and relative to Catholics, those with no religion and other religions are having fewer children over time. The interactions involving the other religious dummy variables were non-significant. That is, although these variables may affect fertility, there is no indication that the effects have grown or declined over time.


**Table 3.** Negative Binomial Models Predicting Self-Reported Fertility over Time.


Note: \* p < .05, \*\* p < .01, \*\*\* p < .001. Standard errors are in parentheses. Coefficients were not transformed by exponentiation for presentation. LR χ<sup>2</sup> = deviancenull - deviancefitted, with the number of parameters for the significance test in parentheses.

The interactions involving age and several of the cohort categories were also statistically significant. The results for age are more complicated and suggest that the inverted U shape seen in Model 3 is "flattening out" over time. That is, in later years, an additional year of age is predicting fewer children than it did in the early period. The results from the Cohort variables echo this conclusion, as people in the later generations are having fewer children than those from the older generations.

### **5. Discussion and Conclusions**

In general, the results presented in this paper confirm many of the findings available in the current literature. For example, we found that marriage increases fertility, as does being black and living in rural areas. On the other hand, additional years of schooling and higher incomes reduce the fertility of respondents in our sample. Importantly, we also found that those who attend religious services more regularly have higher fertility than those who do not attend services as often. This effect, although slightly attenuated, holds true even when religious denomination is controlled for.

The results of our cross-sectional analysis also confirm work that suggests Protestants have fewer children than Catholics. Traditionally, this gap has been attributed to the relatively stronger position on artificial birth control taken by the leadership of the Catholic Church compared to the stance held by the leaders of the various Protestant sects. We also found that Catholics have higher fertility than those who claim "other" religious affiliations and those who claim to have no religious affiliation. This, too, confirms previous research in the area of fertility and religion.

What is unique about our study is the focus on the fertility of the various Protestant sub-denominations. Thanks to work by [9] we were able to separate out the individual Protestant traditions and compare their fertility both to each other and to Catholics. Due to the ideology of some of the more fundamental Protestant sects, particularly conservative Protestants, which closely mirrors the Catholics' views on birth control, we expected that the fertility of conservative Protestants would be significantly higher than that of mainline Protestants and would even rival the fertility of Catholics. That is exactly what we found. When Protestants were separated into four categories (mainline, conservative, Black Protestants, and "other" Protestants) we found that only mainline Protestants reported significantly fewer children than did Catholics, controlling for a host of religious and demographic factors. This suggests that Protestants should not be considered one large homogeneous group. While our results underscored this point with regards to fertility, it is likely that the assumption that Protestants are a homogeneous group will not hold for a number of other outcome measures. For example, it is possible that topics fundamental to the study of religion in society such as religious participation and giving would benefit from studying them in light of the individual Protestant sects.

It is also possible that Catholics are not a homogeneous group. Westoff and Marshall ([38], p. 441) found that Hispanics have higher fertility than other ethnic groups and are more likely to be Catholic. This suggests that within the Catholic religion there may be a difference in fertility along ethnic lines. We leave this topic for future research. Another interesting avenue for future research would be to decompose the "conservative" Protestant group into its various components. That is, using the work of Blanchard *et al.* [39] as a guide, one could examine the fertility of evangelicals, fundamentalists, and Pentecostals separately. It is possible that these groups would show markedly different rates of fertility. As the within-conservative Protestant differences were not the focus of this paper, we also leave that for future research.

In addition to treating Protestants as a heterogeneous group, we also looked at the fertility trends of the various religious denominations over time. For the last few decades, researchers have documented a steady growth in the number of people claiming to be members of religious groups that we would classify as "conservative" Protestant. A number of scholars have attributed this growth to the greater fertility of the people in these groups along with (although to a much smaller degree) conversion and outreach by pastors and laypeople designed to increase membership numbers in specific churches. This argument suggests that the fertility of Protestants should be converging with (growing towards) that of Catholics and that the fertility of conservative Protestants in particular should be diverging (growing away) from that of Catholics.

In general, convergence/divergence implies two possible scenarios: First, the fertility of the specific groups are moving towards each other or away from each other at equal rates over time; that is, one group is having more/less children, while the other is having less/more children. Second, the fertility rates of one group are moving toward the other group, while the rates of the other group remain constant. Third, if two groups are becoming similar over time, then the rates of one group could exhibit a steeper slope than the rates of another group. That is, for example, the rates of Catholics may exhibit slope X, while the rates of conservative Protestants may exhibit a steeper slope, 2X. Thus, the fertility rates of conservative Protestants are "catching up" to the fertility rates of Catholics. It is these ideas that we tested with our trend analyses.

Although not presented in Table 3, we found no evidence for the converging of the fertility of Catholics and the fertility of Protestants. In fact, over time, there was a significant and negative interaction between a dummy variable for Protestant and our linear trend term. This means that the number of children reported by Protestants is growing smaller over time, relative to Catholics. Hence, we found a divergence and not a convergence in our results. As we cautioned above, however, it is probably a mistake to treat Protestants as one large homogeneous group. When Protestants are separated into their respective components and reentered into the model, the effect of time disappears. We found no significant trends in the fertility of mainline, conservative, Black Protestants, or other Protestants relative to Catholics. This means that our second proposition, that the fertility of conservative Protestants is surpassing that of Catholics and growing away from them, was not supported either. It seems as though the fertility rates of the various groups have not changed much over time. The exceptions to this were the fertility of those who claimed "other" religions and those who claimed "no religion." Over time, they both had significantly fewer children, on average, than Catholics.

One of the key implications of these findings is that conservative Protestants are a growing political force in American politics, not only because of the resonance of their movement's message but because of their high fertility rates. The prominence of the "religious right" in the contemporary conservative movement is one reason for its increased political and cultural power. Future research should explore the political implications of conservative Protestant fertility more directly. For example, are children born to conservative Protestant parents more likely to embrace conservative political identifications, vote for conservative candidates, and participate in elections? Additionally, are conservative Protestants' political orientations stable over time, or are there significant age, period, or cohort effects within this group? Conservative Protestant cohorts represent a particularly fruitful avenue for future study. Specifically, one key issue is whether higher fertility rates are translating into greater cohort sizes or, conversely, are a large number of children born to conservatives "leaving the flock." Moreover, it would be interesting to examine the political orientations of conservative Protestant cohorts and whether or not cohort size engenders greater political unity or diversity. Altogether, there is no doubt that conservative Protestants are a major political force in the U.S., however, their relatively high fertility may well ensure that they remain powerful for years to come.

In sum, this paper asked two fairly straightforward questions: 1) Are there differences in the fertility of the Protestant sub-denominations (particularly that of conservative Protestants)? And, 2) Does the growth in the number of conservative Protestants in the United States mean that the gap between Protestant and Catholic fertility is shrinking over time? Our findings suggest that there are indeed differences in fertility among the individual Protestant sects, but that those differences are not growing (or shrinking) over time. Additionally, the gap between Catholics and Protestants is not getting smaller and, especially among mainline Protestants, is getting larger.

### **References**


### **The Connection between Worship Attendance and Racial Segregation Attitudes among White and Black Americans**

### **R. Khari Brown**

**Abstract:** The present study finds that, for Whites, worship attendance is associated with heightened support for racial segregation. This has much to do with the fact that the individuals that attend worship service the least, secular and young adults, tend to be more racially progressive. That is, the extent to which secular and Generation X and Y individuals attend worship services as often as others, worship attendance is associated with weakened opposition to racial segregation. Conversely, worship attendance, religious affiliation, and age cohort are largely unrelated to Black racial segregation attitudes.

Reprinted from *Religions*. Cite as: Brown, R.K. The Connection between Worship Attendance and Racial Segregation Attitudes among White and Black Americans. *Religions* **2011**, *2*, 277–296.

### **1. Introduction**

All major religions promote a common kinship of humankind, in which all people are neighbors and equally valuable in the eyes of the creator. At the same time, religious doctrine has often been used by dominant groups to reinforce hegemony. This dialectic is well exemplified by the Civil Rights Movement where prophetic clergy and lay persons successfully challenged laws and behaviors that, they argued, violated universal and God-given human rights [1-3]. On the other hand, it was not uncommon for White church members to reject the political appeals of their civil rights oriented clergy [2,4,5]. A number of studies suggest that White church members were more heavily opposed to civil rights than were others [4-7]. The conflicting nature of the association between religion and prejudice is also seen in empirical studies that, at times, suggest that worship attendance heightens prejudice among dominant group members while other studies call these relationships into question [6,8,9].

Given that religious congregations are, in many ways, America's central civil societal institution, it is important to investigate their capacity to inform racial attitudes. Nearly two centuries ago, French philosopher Alex De Tocqueville [10] argued that the potential of American religious congregations to inspire independent thought lies in the free space it provides citizens to deliberate in small groups about their roles in protecting and extending their freedoms and opportunities. This is still true today as more Americans are members of, volunteer for, and donate money to congregations than any other non-profit organization [11]. Religious institutions are also one of the top three institutions in which Americans hold a high level of confidence [12,13].

Despite the potential importance of religion in informing American political attitudes, the degree to which worship attendance informs American attitudes about racial segregation has not been well established in survey research. And, the extent to which religious congregations are associated with the willingness of racial/ethnic minorities to live and function within racial/ethnically diverse settings is even less clear. Past studies on religion and out-group attitudes have generally investigated the connection between the dominant group's religion and their tolerance of minority racial/ethnic and immigrant groups [6-8]. While tolerance is a prerequisite for members of diverse social groups to view their life chances as inter-dependent, in isolation, it makes no such assumption [14]. Alternatively, a willingness to live and send one's child to school with members of racial/ethnic out-groups moves one closer to viewing race/ethnicity as an artificial social construct. The current study attempts to fill this gap in the literature by assessing the relationship between worship attendance and racial segregation attitudes among Black and White Americans.

### **2. Religion and Tolerance among Whites**

A majority of Mainline Protestant clergy believe that churches should engage in social justice including reducing racial prejudice, and roughly a fifth of White Mainline, Catholic and Evangelical clergy classify their congregations as maintaining a commitment to social justice [15,16]. Similarly, social justice tends to inform the political behavior of both Mainline and Catholic clergy [15,17]. Congregants are seemingly aware of such discourse, as over half of church-attending Whites report hearing messages about the importance of improving race relations at least once a year in their houses of worship [18]. Over forty percent report that their congregations have hosted or sponsored a program to improve race relations [18].

Nonetheless, few survey and experimental studies find an association between worship attendance and reduced racial prejudice among Whites. Rather, some studies find frequent worship attending Whites--those that attend once a week--to be less prejudiced than individuals that attend between once or twice a year to once a month, but no different than those that never attend [6-8]. Allport [19] explains this relationship by making a case that religious intent informs worship attendance patterns. That is, moderate attendees, those that attend between once or twice a year to once a month, do not truly believe in the core religious tenets of universal love as they attend worship services just enough to satisfy social norms or for self interest reasons. Alternatively, the intrinsically religious attend worship services regularly out of their genuine interest in living a religious life which, in part, emphasizes a universal love of all God's children [19]. However, other studies challenge Allport's [19] argument that distinctions in the religious motivations of the intrinsically and extrinsically religious account for the relationship between worship attendance and attitudes of prejudice. These studies suggest that worship attendance is either unrelated or associated with increased prejudice towards and a desire to maintain distance from racial/ethnic out-groups [6-8,20,21].

The inconsistency in the relationship between worship attendance and racial attitudes may, in part, be linked to the exclusion of secular individuals from many religion and tolerance studies. Given that over three-quarters of secular adults were reared in a religious faith and then became unaffiliated at some point during adulthood implies a questioning of the role that religion plays in their lives and, potentially, of its broader societal importance [6]. Subsequently, secular individuals are more likely than others to question the dominance of a given faith relative to others and the accuracy of religious texts [22,23]. For dominant members, such questioning may lend itself to also questioning social constructions of race/ethnicity historically based upon dominance and marginalization. It follows that secular Whites tend to hold less stereotypical and prejudicial attitudes towards racial/ethnic out-groups than do others [6,8,24]. The fact that only two-thirds of secular individuals report never attending worship services suggests that a substantive number of these individuals are questioning hegemony while attending houses of worship [25]. By excluding such individuals, many religion and tolerance studies are somewhat limited in their ability to explain the connection between worship attendance and racial attitudes. Worship attendance may associate with prejudicial attitudes because secular individuals attend less than do their religiously affiliated counterparts. However, the degree to which secular individuals attend worship services at a similar rate as others may weaken and even reverse the relationship between worship attendance and racial prejudice.

Accounting for cohort effects may add further clarity to the connection between worship attendance/involvement and out-group attitudes. Whites that came of age during the post-civil rights eras of the mid to late 1960s and early 1980s and the Reagan and Bush eras of the early 1980s to the mid 1990s tend to hold less stereotypical attitudes of Blacks [26]. Younger cohorts of Whites also tend to have more positive evaluations of racially integrated neighborhoods and schools and are more likely to support polices that outlaw racial segregation than previous generations [26]. Along these lines, the Religion and Politics Study [22] suggests that younger Whites are more likely than older Whites to be concerned about racial discrimination. Like secular Whites, younger White cohorts that question racial inequality may also raise larger questions about hegemony which includes looking to dominant religious institutions as the primary source for moral instruction. Moreover, in the same way that younger cohorts of Whites tend to question the basis of racial inequality, they are also more likely than others to question the inerrancy of the Bible, believe that all religions contain some truth, and that all religions are equally good ways to relate to God [22]. The fact that young adults attend worship services less often than do older cohorts may partially explain why worship attendance rarely contributes to racial progressive attitudes among Whites.

### **3. Religion and Tolerance among Black Americans**

It is unlikely that cohort effects, worship attendance or religious affiliation have the same impact on Black racial attitudes as it does for Whites. Given that marginalized groups tend to be restricted from opportunity structures, they are unlikely to face the same moral dilemma over the extent to which institutions should allow for increased opportunities for all social groups. The reasoning follows that because Blacks are socio-economically disadvantaged, racial integration tends to reduce concentrations of these groups in communities with few social economic resources in the form of jobs, quality education, health care facilities, and other qualities of life [27-29]. The socioeconomic benefits of racial integration for racial/ethnic minorities likely contributes to Blacks of disparate age groups and religious affiliations maintaining relatively high levels of support for racial integration. Such an outlook may, in part, explain why Black clergy and laity were overrepresented in civil rights demonstrations during the 1950s and 1960s [1,2]. Relative to Whites, there was also much greater support among Blacks for open housing campaigns and civil rights efforts more generally [25].

The vast majority of Blacks continue to support racial integration and policies that encourage neighborhood and school racial/ethnic diversity [30,31]. In addition, Blacks remain more supportive than Whites of religious institutions participating in political movements and in fighting poverty and Blacks are more supportive than Whites of religious institutions fighting racial discrimination [22]. Such attitudes may explain why Blacks are more likely than Whites to hear sermons about poverty and why Blacks are more likely than Whites to hear sermons about racial discrimination [22]. Along these lines, Black congregations are over-represented in faith-based community organizing firms that are committed to improving the quality of life within poor communities via interracial grassroots political coalitions [32]. Moreover, while secular and younger individuals may partially explain the relationship between worship attendance and racial attitudes among Whites, this is not likely the case for Blacks. This leads to the following research question: to what extent does religious affiliation and age cohort explain the relationship between worship attendance and racial segregation attitudes among White and Black Americans?

### **4. Sample**

This study relies upon the 1972-2008 General Social Survey (GSS) cumulative data file. In most years since 1972, the National Opinion Research Center conducted the GSS via face to face interviews, computer-assisted personal interviews, and telephone interviews [33]. The GSS is a national representative sample of adults 18 years and older living in non-institutionalized settings within the U.S. [33]. This study is the largest project receiving funding from the National Science Foundation's Sociology Program. Outside of the U.S. Census, this data is more heavily analyzed than any other source of social science data. For the purpose of replication, many of the demographic and attitudinal questions have remained constant since 1972 to allow for longitudinal studies and the replication of previous results. A number of studies also contain questions of special interests, such as religion in 1991, 1998, and 2008. And, while the exact wording for some questions has changed from survey to survey, the cumulative data file retains consistency across surveys [34]. In total, the 1972-2010 GSS has roughly 5,400 variables, time-trends for nearly 2,000 variables, and 257 trends with over 20 data points.

Up until 1994, 1,500 was a typical sample size. After 1994, the GSS became biennial and sample sizes increased to 3,000. Response rates vary between 74 and 82 percent [33]. Depending upon the years in which the dependent variable questions were asked, the Black sample ranges from 263–2,173 and the White sample from 1,685–16,468.

### **5. Measures**

### *5.1. Dependent Variables: Racial Segregation Attitudes*

To get as complete a picture as the General Social Survey allows on Americans' racial segregation attitudes, this study relies upon a number of racial segregation variables asked between 1972 and 2008. Between 1972 and 2008, opposition to open housing was assessed by the extent to which respondents would support a policy that allowed them to decide to whom they are willing to sell their house, even if they preferred not to sell to (Negroes/Blacks/African-Americans). Between 1972 and 1996, support for homeowner discrimination was assessed by the extent to respondents believed that White people have a right to keep (Negroes/Blacks/African-Americans) out of their neighborhoods if they want to, and (Negroes/Blacks/African-Americans) should respect that right. Between 1988 and 2008, opposition to living near Blacks was assessed by the extent to which respondents were opposed living in a neighborhood where half their neighbors were Black. This same question was asked about Whites between 2000 and 2008. Between 1972 and 1996, opposition to sending one's child to school with [Whites/(Negroes/Blacks/African-Americans)] was assessed by the extent to which respondents were opposed to the idea of sending their child to a school in which half of the children are [Whites/(Negroes/Blacks/African-Americans)]. Between 1972 and 1996, opposition to busing was assessed by the extent to which respondents opposed the busing of (Negro/Black/African-American) and White school children from one school district to another to integrate the schools. The varying years in which these racial segregation variables are included in the analyses presented below are based upon the years in which the GSS presented those questions on their survey.

### *5.2. Independent Variables: Worship Attendance, Religious Affiliation, and Age Cohort*

### 5.2.1. Religious Affiliation and Worship Attendance

This study relies on Steensland *et al.*'s [35] classification of religious denominations. Membership status in national religious organizations such as the National Council of Churches and the National Association of Evangelicals are used to classify various Baptist, Methodist, Lutheran, Presbyterian, and Episcopalian denominations into Evangelical, Mainline, and historically Black Protestant traditions. Respondents were divided into the nominal categories of Evangelical, Mainline Protestant, Catholic, Jewish, Black Protestant, Other Protestant, Other-Faiths, and Secular. The secular category includes those who do not identify with or affiliate with a religion. Because of sample size considerations, there are slight differences in the denominations/faiths represented among the varying racial/ethnic groups included in this study. There are no separate dummy variables of Black Protestant Churches in the White American analyses. Jewish and other Protestant dummy variables are not included in the African American analyses. Such individuals along with followers of many other faiths are instead grouped in an "other faith" category. This study relies upon a standard worship attendance variable that assesses the frequency to which individuals attend houses of worship on a scale that ranges from 1—never attending, to 8 attending more than once a week.

### 5.2.2. Age Cohort

This study borrows Schuman *et. al*.'s [26] coding of age cohorts in which the youngest persons included in the pre-civil rights age cohort are persons that reached age 18 in 1953, a year before the Brown *vs.* Board of Education Supreme Court Decision. The civil rights cohort includes those individuals that came of age during the successful civil rights struggle between 1954 and 1965. The post civil rights cohort came of age during the mid 1960s to early 1980s (1966-1980), Generation X came of age between 1981 and 1995, and Generation Y came of age after 1995. Because small samples of individuals from Generation Y were asked many of the racial segregation questions, they were grouped with Generation X to form the Generation XY cohort.

### *5.3. Control Variables*

The current study controls for the standard demographic factors of college education, family income, gender, children in the household, party identification, and living in the South.<sup>7</sup> This study also takes into consideration the years during which the racial segregation attitude questions were asked.8 

### **6. Results**

### *6.1. Religion, Age Cohort, and Racial Segregation Attitudes among White Americans*

In all cases but one, the analyses presented in Tables 1 through 3 suggest that both religious affiliation and age cohort serve as lurking variables that impact the relationship between worship attendance and White racial segregation attitudes. Because secular and younger Whites attend houses of worship less often and are more likely to oppose racial segregation than others, worship attendance, on average, contributes to heightened support for racial segregation. However, the extent to which secular and younger Whites attend at the same rate as others, attendance contributes to increased opposition to racial segregation to the point that previous effects are nullified or reversed such that attendance reduces opposition to racial segregation. In addition, the likelihood ratio tests suggest that the addition of religious affiliation and age-cohort significantly improves the fit of the models presented in these analyses.

Table 1 indicates that religious affiliated and older cohorts of Whites are more likely than others to attend worship services. The reduced model of the opposition to open housing analyses suggests that worship attendance heightens opposition. However, taking into account age-cohort weakens the relationship between worship attendance and open housing attitudes to the point that worship attendance is no longer associated with such preferences. Although worship attendance still heightens opposition to open housing in the religious affiliation model, accounting for religious affiliation weakens this relationship. The full model indicates that worship attendance is unrelated to open housing attitudes. These analyses also indicate that the pre civil rights, civil rights, and post civil rights cohorts are more opposed to open housing than are generation XY cohorts. In addition, Evangelicals, Mainliners, and Catholics are more likely than are secular Whites to oppose open housing policies.

The reduced model of the support for homeowner discrimination analyses, also reported in Table1, suggests that worship attendance heightens support. However, taking into account age-cohort and religious affiliation weakens this relationship to the point that worship attendance is unrelated

<sup>7</sup> Missing values for family income, age cohort, and church attendance were imputed from an imputation procedure that organizes missing cases by patterns of missing data so that the missing-value regressions can be conducted efficiently. The imputations did not significantly or substantively alter the analyses.

<sup>8</sup> These analyses are weighted to account for non-respondents within the sampling design [33].

to support for homeowner discrimination in the age cohort and religious affiliation models. The full model indicates that worship attendance actually reduces support for homeowner discrimination. These analyses also indicate that the pre civil rights, civil rights, and post civil rights cohorts are more supportive of homeowner discrimination than are generation XY cohorts. In addition, Evangelicals, Mainliners, and Catholics are more supportive than are secular Whites.

The reduced model of the opposition to living near Blacks analyses reported in Table 2 suggests that worship attendance is unrelated to such preferences. Worship attendance nearly reduces such opposition when age cohort and religious affiliation are respectively accounted for in the agecohort and religious affiliation models. When both religious affiliation and age cohort are accounted for in the full model, worship attendance does reduce opposition to living in integrated neighborhoods with Blacks. These analyses also indicate that the pre civil rights, civil rights, and post civil rights cohorts are more opposed than are generation XY cohorts. In addition, Evangelicals, Mainliners, Catholics, Jews, and other Protestants are more opposed than are secular Whites. Contrary to the other cases, Table 2 also reports that age cohort does not explain the relationship between worship attendance and opposition to living near Hispanics. These analyses also indicate that Evangelicals are more likely than are secular Whites to oppose living near Hispanics.

The reduced model of the opposition to school integration reported in Table 3 suggests that worship attendance is unrelated to such preferences. Taking into account age-cohort and religious affiliation strengthens this relationship to the point that worship attendance reduces opposition in both the age cohort and religious affiliation models. In the full model, worship attendance continues to reduce opposition to school integration. These analyses also indicate that the pre civil rights, civil rights, and post civil rights cohorts are more opposed to school integration than are generation XY cohorts. In addition, Evangelicals, Mainliners, Catholics, Jews, and Other Protestants are more opposed than are secular Whites. The reduced model of the opposition to busing analyses, also reported in Table 3, indicates that worship attendances increases opposition. Although worship attendance continues to heighten opposition in both the age cohort and religious affiliation models, the relationships are weakened. In the full model, worship attendance is no longer associated with opposition to busing. These analyses also indicate that the pre civil rights, civil rights, and post civil rights cohorts are more opposed to busing than are generation XY cohorts. In addition, Evangelicals, Mainliners, Catholics, and Other Protestants are more opposed than are secular Whites. The predicted probability estimates listed in the appendix further illustrates that worship attendance reduces support for racial segregation attitudes as age cohorts and religious affiliation are taken into account. Overall, Tables 1 through 3 also suggests that college graduates, non-southerners, women, and individuals that were interviewed more recently tend to oppose racial segregation.



9 The Likelihood Ratio (L.R.) Test represents the change in the log likelihood between the reduced and nested models.

Mainline --- --- 0.557\*\*\* 0.439\*\*\* --- --- 0.424\* 0.386

(0.105) (0.107) (0.208) (0.210)

Catholic --- --- 0.533\*\*\* 0.484\*\*\* --- --- 0.328 0.332

(0.104) (0.105) (0.211) (0.212)





**Table 4.** *Cont.*




### *6.2. Religion, Age Cohort, and Racial Segregation Attitudes among African Americans*

The analyses presented in Tables 4 and 5 largely suggest that, for Blacks, age cohort and religious affiliation play a very limited role in explaining the connection between worship attendance and racial segregation attitudes. It appears that religious affiliation and age-cohorts weaken the relationship between worship attendance and support for homeowner discrimination among Blacks in Table 4. It also appears that religious affiliation strengthens the relationship between worship attendance and opposition to busing such that worship attendance weakens opposition to busing. In all other cases, religious affiliation and age-cohort do not further explain the association between worship attendance and Black racial segregation attitudes10. Age cohort and religious affiliation are fairly inconsistent predictors of Black racial segregation attitudes. Demographic factors also do a poor job in predicting such attitudes among Blacks.

### **7. Discussion and Conclusion**

This current study suggests that the connection between worship attendance and White racial attitudes is largely a function of religious affiliation and age-cohort. For Whites, worship attendance is associated with more conservative racial attitudes because the religious affiliated and older cohorts attend more often and are more supportive of racial segregation than are secular persons and younger cohorts. Moreover, when the worship attendance gap is eliminated between secular and religious affiliated Whites and between younger and older Whites, attendance weakens support for racial segregation. These findings are consistent with research that suggests that secular and younger adults maintain more tolerant and progressive attitudes towards out-groups [6-8]. The same desire to understand the meaning of life that drives such individuals to question religious doctrines also likely fuels a desire to understand constructions of race/ethnicity that contributes to separation and conflicting interests between Whites and non-Whites. As such, the presence of secular and younger age cohorts of Whites within houses of worship are key to understanding the degree to which worship attendance is associated with support for racial segregation in residential and school contexts.

These findings may suggest that the attitudes attendees bring to their worship services reinforce existing racial attitudes. This is not to suggest that clergy have no influence over congregant racial attitudes. A number of studies suggest that clergy are capable of influencing congregant political attitudes and ideologies [36-40]. At the same time, clergy are particularly sensitive to member preferences as religious congregations are voluntary associations that are almost completely dependent upon congregants for money, gifts, and volunteers [4,5]. As such, even in instances in which Mainline, Evangelical, and Catholic clergy are more racially progressive than their congregants, they are unlikely to push too hard against entrenched attitudes. At this point, however, claims about the capacity of congregants relative to clergy to inform racial attitudes are merely

<sup>10</sup> Because the likelihood ratio tests were non-significant in all but 1 model within the Black analyses, it is not included as part of their analyses in Tables 4-5. For the same reason, the probability estimates for Blacks are not included in the appendix.

speculative. The present study is not able to assess the degree to which congregants are exposed to discussions about race from; clergy, other religious leaders, congregants, and/or if they are taking part in such discussions in their houses of worship. Future research in this area is necessary to make more definitive claims about how the source of race discourse within houses of worship may inform the connection between worship attendance and White racial attitudes.

Conversely, for Blacks, attendance of worship services, religious affiliation, and age cohorts are largely unrelated to racial segregation attitudes. Such findings are understandable given that Blacks are under-represented among the American middle class and over-represented among the poor. Moreover, it is conceivable that both worship going and non-worship going blacks have a compelling interest to support racial/ethnic integration as a means to improve their individual and group life chances. This may explain why worship attendance is largely unrelated to Black racial segregation attitudes. However, at this point, the provided explanation serves, again, as only speculation. Further research is required to determine if perceptions of racial inequality and opportunity structures largely explain the connection between religion and racial attitudes among Blacks. That being said, these findings are consistent with Brown's study that found religious and non-religious Blacks to maintain similar positions on the importance of racism and economic barriers in explaining racial inequality [41].

In sum, for Whites, questions about dominance and marginalization among the secular and young adults likely contribute to their more progressive racial attitudes. The fact that these groups also tend to possess more critical attitudes about religious institutions likely contributes to their lower attendance rates, which, in part, explains why worship attendance is associated with increased support for racial segregation among Whites. Alternatively, the fact that racial integration extends social-economic opportunities to Blacks likely contributes to age cohort, religious affiliation, and worship attendance maintaining a limited relationship with their racial segregation attitudes.

### **References**



**Table A1.** Predicted Probability Estimates of the relationship between Worship Attendance and White Racial Segregation Attitudes: Probability Estimates are based upon analyses within Tables 1-3. 


### **Marital Naming Plans among Students at Four Evangelical Colleges**

### **Kevin D. Dougherty, Melanie Hulbert and Ashley Palmer**

**Abstract:** Despite increasingly egalitarian gender roles in the United States, when the wedding bells ring for heterosexual couples, husband and wife still commonly emerge sharing the man's last name. Largely missing from previous studies of marital name change is the influence of religion. We examine the marital naming plans of 199 students from four Evangelical colleges. Nearly all these students planned to marry and more than 80% planned to follow the traditional naming pattern for their gender. Bivariate correlations and logistic regression models reveal that private prayer and more literal views of the Bible correspond to plans for a traditional marital surname. Yet, only a small minority of students evoked religious language to justify their surname choice. Gender roles, identity, and tradition were dominant themes in their explanations. Whether recognized or not, personal religiosity and the model of marriage cultivated in religious families guide the marital naming intentions of Evangelical students. Thus, religion operates as an invisible influence shaping ideals of marriage and family within Evangelical subculture.

Reprinted from *Religions*. Cite as: Dougherty, K.D.; Hulbert, M.; Palmer, A. Marital Naming Plans among Students at Four Evangelical Colleges. *Religions* **2014**, *5*, 1116–1131.

### **1. Introduction**

"It is my privilege to introduce to you for the first time, Mr. and Mrs. [insert male name here]." These are the closing words of countless church weddings for generations in the United States. Although, like church weddings, the practice of U.S. women taking their husband's last name remains common, it is now far from universal.

Marital naming conventions hold important implications for individuals and societies. Surnames trace descent and speak to patterns of familial and societal authority. Changing attitudes toward marital surnames in heterosexual marriages is a subject of substantial research. Known correlates to progressive marital naming views include gender, ethnicity, socioeconomic status, and gender ideology. Remarkably little previous research delves into the influence of religion and religious contexts in shaping marital naming choices. It is a striking omission given the association of religion to issues of gender, marriage, and family [1–4].

The purpose of this study is to explore marital name change plans among students at four Evangelical Christian colleges. Evangelicals have been on the front lines of culture war battles over marriage and family [5]. Local churches and parachurch organizations like Focus on the Family are recognizable proponents of Evangelical ideals of marriage and family. Evangelical colleges and universities are foundational to the Evangelical subculture. While surveys of college students on the topic of marital name change are common, no previous research looks within Evangelical colleges specifically. We extend prior research by concentrating on the marital naming plans of students in Evangelical colleges and examining a wider array of religion variables.

### **2. Marital Naming Patterns**

The social expectation placed on women to take their husband's surname has a long history. Until overturned by a Supreme Court ruling in 1975, many states had laws that required women to adopt their husband's surnames at marriage. Even without a legal requirement, the adoption of a husband's surname remains the most common naming option for brides in America [6–10]. More than nine out of ten U.S. women take their husband's last name at marriage [7,8]. Nearly three-fourths of American adults agree that it is generally better if a woman changes her name when married and half believe that marital name change for women should be legally required [11]. Even college students show remarkable consistency over the past two decades in their marital naming intentions [12].

Varied explanations have been put forward for why American women continue to take their husband's last name. Some see it as a drift toward more conservative values in the society as a whole [13]. Others contend that empowered to choose a surname at marriage many women choose their husband's surname as a matter of convenience or tradition [14,15]. In essence, the maiden name is no longer a political issue. One journalist memorably summarized the decision-making of contemporary women: "[W]hich name do you like the sound of? What do you feel like doing? The politics are almost incidental…in a mundane way, having the same name as your children is easier." [16].

Over the past two decades, social scientists strived to understand the surname choices people make, particularly women. Gender, ethnicity, education, and gender role ideology consistently appear as influences on marital naming choices [7,8,10,11,14,17–22]. Only a handful of studies on marital naming considers religion, and even fewer employ more than one or two religion variables. These studies yield mixed results. In a random sample of adults from one Midwestern state, church attendance was negatively associated with tolerance for a woman keeping her maiden name at marriage [20]. In other studies, church attendance showed no relationship to marital naming choices [8,9]. Similar conflicting findings surround religious traditions. Some find Catholics to favor traditional marital surnames [15,18], whereas others find no difference among Catholics, Protestants, and persons with no religious preference [20]. In one other study, people who favored more literal readings of the Bible held more traditional attitudes toward marital name-change, even when controlling for numerous sociodemographic variables [11]. No study has tested whether literal views of the Bible actually predict marital naming choices however. Furthermore, none of the aforementioned studies have focused specifically on Evangelicals, despite the cultural and political prominence of this segment of American Christianity.

### **3. Gender and Marriage among Evangelicals**

Evangelicals, also referred to as Conservative Protestants, often are seen as vocal proponents of traditional views on marriage and family. Evangelicalism is a branch of Protestant Christianity that emphasizes salvation through a personal relationship with Jesus Christ, the authority of Biblical scripture, and the centrality of evangelism. More than a quarter of American adults are counted as Evangelicals [23,24]. A dominant discourse within Evangelicalism is an essentialist view of gender in which men and women are inherently different [25]. Men and women are understood to inhabit separate spheres in social life, with men filling public and provisionary roles and women overseeing domestic responsibilities. This language of divinely-ordained gender differentiation features prominently in the Conservative Protestant dialogue on gender and family [2,25,26]. It is a view maintained and reinforced through places of worship, religious education, publications, and relationships with co-religionists [2,25,27,28].

Correspondingly, compared with adherents to other religious traditions, members of Conservative Protestant denominations report more traditional views of gender, including support for a patriarchal family structure [29] and agreement that there are proper roles for men and women in society corresponding to the public and private spheres [30,31]. Gender role ideology for Evangelicals is couched in a language of headship and submission in which leadership and authority in the family are ascribed on the basis on gender. These concepts, however, are not interpreted uniformly—traditionalists endorse the patriarchal order of *wifely* submission, while other Evangelicals use the language of *mutual* submission to describe marriage as a non-hierarchical partnership [32].

Despite these ideological distinctions, for both traditionalist and egalitarian Evangelicals, family life operates according to a pragmatic egalitarianism in which decision-making and domestic work are a function of who has the most time, opportunity, and expertise in a particular area [2]. Although there are exceptions to this pragmatism—chiefly in childrearing—utilizing a language of submission and headship is largely symbolic. By assigning familial headship and authority on the basis of gender, Evangelicals on either side of the divide simultaneously identify themselves with the values of a distinctive religious subculture and solidify men's role in the family. Subject to the same economic shifts as other Americans, most Evangelical families are now dual-earner couples that depend on the economic provision of both the male and female partner to support the household. Enacting a traditional script ensures that men continue to play a central role in the family through spiritual provision and leadership even as their role in breadwinning wanes [33].

Given these findings, how might young Evangelicals view and talk about marital naming choices? Based on the traditional gendered scripts prominent in Evangelicalism, we expect marital naming plans among students in Evangelical colleges to reflect this tradition. Specifically, we anticipate that both young men and young women on these campuses will favor the male surname for marriage and that religious influences will shape surname choice.

### **4. Research Setting**

College campuses are popular sites for studying marital naming patterns [12,19,22]. College serves as a formative setting in which a young person sharpens long-term goals and preferences related to work and family [34]. This maturation process entails the development of attitudes, beliefs, and values regarding gender roles [35]. Young adults' place in the life cycle—prior to marriage and childbearing—and the identity formation that takes place during the undergraduate years together provide a unique opportunity to analyze marital naming choices as they reflect the social and environmental factors that influence personal aspirations and expectations.

Young adulthood is also a time of religious transition for many marked by a decline in aggregate levels of religious belief and practice [36]. Evangelical colleges stand out as a startling contrast to this developmental trend. Students at Evangelical colleges show high levels of religiosity [37,38]. Attitudes toward marriage and family in these settings likewise favor tradition. In a study of nine U.S. Evangelical colleges, researchers found "a sizable percentage of students continue to support traditional concepts of marriage and female roles" ([37], p. 95). Over half of those sampled agreed that a husband has primary responsibility for the spiritual wellbeing of the family; more than twothirds agreed that a woman should put her husband and children ahead of her career. Among female students surveyed at a separate Evangelical college in the Midwest, the majority preferred a complementarian view of gender—a position that advocates male headship and female submission as complementary roles established by God for the operation of family and society [39]. While this ideology argues for equality of worth between men and women, it states that men and women have differing and unequal roles. Hence, Evangelical colleges, like Evangelical churches, are an organizational context in which ideas about gender, marriage, and family get reinforced. We step inside this organizational context to examine the marital naming plans of students.

### **5. Method**

We surveyed students enrolled in introductory sociology courses at four colleges affiliated with the Council for Christian Colleges and Universities (CCCU), an international association of 119 intentionally Christian colleges and universities. To belong, schools must have a Christian mission, Christian curricular and extra-curricular programming, and hire only professing Christians as faculty and administrators. We partnered with sociology faculty at four CCCU schools: one in the West, one in the Midwest, one in the South, and one in the Northeast. The schools ranged in size from 1500 to 3400 students. The theological and denominational backgrounds of all four colleges place them squarely within the Evangelical Protestant tradition.

A web-survey was administered between 7 February and 5 March 2011. The survey had questions on marital naming plans, religion, family background, and demographics. Cooperating faculty sent the survey URL to students in their introductory sociology courses. At our prompting, faculty sent two email reminders to students roughly two weeks apart. The survey was restricted to unmarried students aged 18 and older. In the end, 199 students completed the survey. Because these students represent a convenience sample, we are unable to generalize to any larger population. Nevertheless, this snapshot of students permits us to examine perceptions of marriage and marital roles in a unique religious context. In keeping with prior research and given the prevailing values of the Evangelical subculture, our study centers on monogamous heterosexual marriages.

Our dependent variable comes from the question: "If you marry, do you plan to…" Response options included taking a spouse's last name, hyphenation, keeping one's last name, taking a spouse's last name while keeping a birth name as a middle name, other, and "I do not plan to marry." Our analysis begins with a simple frequency distribution by gender for all of these response options. For subsequent analyses, we created a dichotomous variable to distinguish respondents who plan to follow the traditional naming pattern for their gender (*i.e.*, women who plan to take their husband's last name and men who plan to retain their surname from birth, coded 1) from those who do not (coded 0).

We employ a wider range of religion variables than considered in previous research on this topic. We inquired about religious service attendance at age 12 and at present. The first is a measure of religious socialization. The second is a common marker of public religiosity. Both variables had a nine-point scale from 1 = never to 9 = several times a week. Two other variables pertain to private religiosity. The survey asked respondents, "About how often do you pray?" There were seven response options, which we coded from 1 = never to 7 = several times a day. Another question asked, "Outside of attending religious services, about how often do you read the Bible?" We treated these options in similar fashion with 1 = never to 9 = several times a week or more. In addition to variables of religious practice, we include measures of religious belief and selfperception. A customary belief measure is biblical literalism. Students selected a statement indicating their personal beliefs about the Bible from a list of four ordered options ranging from "The Bible is an ancient book of history and legends" to "The Bible means exactly what it says. It should be taken literally, word-for-word, on all subjects." Higher values indicate more literalist orientations. Finally, to gauge respondent's perception of their religiosity, we asked, "How religious do you consider yourself to be?" Students chose from four options: not at all religious (coded 1), not very religious (coded 2), somewhat religious (coded 3), and very religious (coded 4).

Demographic and family background characteristics serve as control variables in our statistical analysis. Building on previous research on marital naming choices, we control for gender (1 = female), race (1 = white, Non-Hispanic), age (in years), college, mother's last name (1 = mother had same last name as father), mother's level of education (from 1 = less than high school diploma to 5 = graduate degree), mother's employment status during respondents' childhood (1 = homemaker), and parents' marital status (1 = divorced). In addition to these customary control variables, we add two others. We test perceived family expectations regarding martial names (for a comparable variable in a study of offspring surnames, see [40]). A question on the survey asked: "If you marry, what does your family expect you to do?" Response options included taking a spouse's last name, hyphenation, keeping one's last name, taking a spouse's last name while keeping a birth name as a middle name, and other. We created a dichotomous variable to differentiate respondents who believe their family expects them to follow the traditional naming pattern for their gender (coded 1) from those who do not (coded 0).

Respondents were 73% young women and 27% young men. The majority was White (72%). As would be expected for introductory courses, students tended to be young. The average age was 19.59. The number of respondents varied considerably across the four CCCU colleges due to differences in the number of introductory sociology classes being taught and class sizes. Over half (56%) were from the Northwest CCCU college, 24% were from the Midwest college, 11% were from the South college, and 9% were from the Northeast college. Respondents were religious. Median values indicated that they attended religious services weekly, read the Bible about weekly, and prayed daily. They held orthodox views about the Bible as well. Three-fourths (74%) believed that "the Bible is perfectly true, but should not be taken literally, word-for-word", while 17% supported the view that the Bible "should be taken literally, word-for-word, on all subjects". In terms of self-identification, 36% of these students considered themselves very religious and another 46% described themselves as somewhat religious. By and large, respondents came from traditional families. Less than a quarter (22%) had divorced parents. For 35%, mom was at home for part or all of their childhood. The marital surnames of parents followed traditional gender scripts. Nine out of ten respondents (88%) had mothers who took her husband's last name. Equally telling, 83% of respondents believed that their families expect them to follow the same tradition if they marry.

We present our findings in three stages. First, we describe the marital naming plans of female respondents and male respondents in our sample, and we compare these stated intentions to the results of previous studies based upon students at non-religious colleges. Second, we test the relationship between religion and martial naming plans using bivariate and multivariate analysis. We report correlation coefficients and logistic regression models, which regress plans to follow the traditional naming pattern (yes/no) on religion variables and control variables. Third, we consider students' explanations for their martial naming plans. An open-ended survey question asked respondents to explain in their own words their planned naming choice: "Why do you plan to use the last name you do, if you marry? Briefly explain." One hundred and eighty-nine respondents (95% of the total sample) did. We conclude our analysis with thematic coding of these qualitative responses.

### **6. Marital Naming Plans**

Table 1 displays the marital naming plans by gender for our sample. Among female students, 81.4% reported that they would change their last name to that of their spouse. A hyphenated name, cited by 9.0% of female respondents, was the most prevalent alternative. Approximately five percent said they would take their spouse's last name and keep their family surname as a middle name. Another 3.4% of female respondents selected "other". A text-box in the survey allowed respondents to clarify what other naming option they would pursue. Several female respondents indicated that their choice of a marital surname would depend on their spouse's last name or how her first name would sound with his last name. Only one female student planned to keep her surname from birth. One other female student said that she did not plan to marry.


**Table 1.** Marital Naming Plans of Respondents.

The young men in our sample were spread across fewer response options. Eighty-five percent of male respondents planned to retain their surname from birth when married. Seven percent planned to hyphenate their name. An equal percentage said they would change their last name to that of their spouse. Not a single male student in our sample planned to use his last name as a middle name or use some other surname option. All planned to marry.

We see in Table 1 the high emphasis placed on heterosexual marriage in the Evangelical subculture. Only one student out of 199 selected "I do not plan to marry". When they do marry, over 80% of men and women in our sample planned to stick with marital naming tradition. Male respondents and female respondents expect to share the husband's last name. There was no statistically significant difference between men and women in this regard (Chi-Square = 0.39, *p* = 0.53).

It is difficult to know whether Evangelical students in our sample are any more or less likely to follow traditional marital naming customs than are students at non-religious colleges and universities. No national data on college students' naming intentions exist. The best that we can do is to compare our findings to previous studies conducted at other colleges and universities. In a 1990 random sample of students at a small residential college in the Midwest, 81.6% of female students planned to take their husband's last name, if they married, and 7% planned to hyphenate their name [19]. These percentages are very close to the percentages for female students in our sample, in which 81.4% planned to take their husband's last name and 9.0% planned to hyphenate. Male students at the same Midwestern college and at a large Eastern university were even more prone to traditional marital naming [12]. In 1990, 97.9% of men interviewed at the small Midwestern college planned to keep their family surname, if they married; sixteen years later on the same campus, 96.9% of men interviewed gave the same response [12]. A 2006 convenience sample of 369 students from a large Eastern university found that 100% of male respondents planned to keep their birth surname, if they married [12]. By comparison, 85.2% of men in our sample intended to keep their birth surname when married. It is important to note that neither the small Midwestern college nor the large Eastern universities were identified as religious. While it is risky to make comparisons across samples collected in different ways and at different points in time, the limited evidence that we do have counters our expectation that Evangelical students are dramatically more traditional in their marital naming plans. When it comes to marital surnames, tradition seems to be popular on religious and non-religious campuses. Next we consider the role of religion in upholding marital naming traditions on Evangelical campuses.

### *6.1. Religious Influences*

Table 2 is our first look at potential religious influences on marital naming choices. We test for significant zero-order correlations for each religion variable in relation to a respondent's plan to follow with the traditional naming pattern for her/his gender. In this highly religious sample, church attendance is not significantly correlated with plans for a traditional marital surname. Neither is private Bible reading. The only religious practice variable that is significant is prayer (*r* = 0.18, *p* < 0.05). Respondents that pray more frequently are more likely to say that they plan to follow the marital naming convention for their gender. Biblical literalism is also significant (*r* = 0.21, *p* < 0.01). More literal views of the Bible correlate with taking a traditional marital surname. Self-rated religiosity (*r* = 0.20, *p* < 0.01) is significantly correlated with plans for a traditional marital name as well.


**Table 2.** Religious Influences on Plans to Follow Traditional Marital Naming Pattern.

\* *p* < 0.05, \*\* *p* < 0.01 (Pearson's Correlation Coefficients).

Table 3 tests the influence of religious variables on plans for a traditional marital name when controlling for family background and demographic characteristics. We build upon our bivariate findings by testing only the religion variables that had a significant zero-order correlation with traditional marital naming plans. We test the influence of prayer, Biblical literalism, and self-rated religiosity in separate models, since these three religion variables are significantly correlated with one another and including all three in the same model would violate a regression assumption. We present our results as odds ratios. Values greater than 1.0 indicate increased odds of following a traditional naming pattern, while values less than 1.0 denote reduced odds of following this pattern. Model 1 regresses plans for a traditional marital name upon prayer and control variables. More frequent prayer is associated with a greater likelihood of following a traditional marital naming pattern (odds ratio = 1.39, *p* < 0.05). In Model 2, Biblical literalism shows a similar effect. More literal views of the Bible significantly and strongly predict support for a traditional marital surname (odds ratio = 2.99, *p* < 0.05), holding constant other family and demographic influences. Adding the control variables renders self-rated religiosity non-significant. The correlation between selfrated religiosity and traditional marital naming plans observed in Table 2 seems to be explained away by other characteristics.

Most of the family and demographic control variables are not significantly related to traditional marital naming plans in our sample. Mother's education, mother's employment, divorced parents, gender, race, and age were not statistically significant in any model of Table 3. Differences in martial naming plans across the four colleges also did not appear significant in Table 3. Two control variables did stand out. Respondents whose mother took her husband's last were over five times more likely to say they would follow the same tradition in their marriage. An even stronger relationship existed for family expectations and plans for a traditional martial surname. Respondents who believed that their family expected them to adhere to traditional marital naming customs were 17 to 20 times more likely to state that would follow this pattern.


**Table 3.** Religious, Family, and Demographic Influences on Plans to Follow Traditional Marital Naming Pattern.

Logistic regression odds ratios; a Contrast category is South CCCU college; \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

### *6.2. Student Explanations*

Three dominant themes emerged for why most students intend to adhere to marital name change custom: gender roles, identity, and tradition. Although only five students (one man and four women) mentioned God or the Bible directly, students' comments illuminate the foundation religion provides for guiding individual choices. We describe each theme and provide representative examples below.

Gender roles stood out as a prevalent theme guiding students' marital naming plans. Nearly a quarter of students referenced gender roles in their explanations. Men and women are inherently different, respondents contended. Furthermore, the success of marriages and societies depends on men and women playing their gender-specific roles. These ideas about distinct male and female roles align with a complementarian gender ideology, which Colaner and Warner found to be prevalent in another Evangelical college [39]. Illustrated in the following comments, male students claimed a dominant position, ordained by God, as the provider in marriage, while female students championed their future role as the devoted and dutiful wife.

"Because I am the man in the relationship. God has placed me at the head of the household and it's a huge role. I think it is important for the woman to take the last name of the man because it shows that the two of them have become one." (male student)

"The man is the head of the household. Woman in the Bible aren't [sic] mentioned in Genesis that often because the man carries on the last name of the family. When it comes down to it a name is just a name, but not taking your husband's last name is in a way disrespectful and not submissive." (female student)

"I plan to use my last name because I am a man and the woman taking the name of the man in marriage is a sign of her acceptance of her role in marriage." (male student)

Identity was a dominant theme expressed by about a third of students, both male and female. However, different from the idea of preserving individual identity or creating a new shared identity as described above, these respondents linked identity to traditional gendered expectations surrounding marriage. The basis of identity differed by gender. Young men tied personal identity to their family of origin. Extending their family name (*i.e.*, patrilineal descent) was important to them, as we heard in responses such as these:

"My last name means a lot to me and my family. Me, being a male, I have the privilege to pass down my name to my kids and also my wife." (male student)

"Because my name means a lot to me and it is part of my identity and I would like to keep my name going. I also believe it shows who I am." (male student)

These comments draw attention to a patriarchal orientation that assumes male lineage and masculine identification. The explanations of these young men reflect the way in which patriarchy both shapes and normalizes their choice to keep their name upon marriage.

We heard something different from the young women in our sample. For many, identity after marriage was defined in relation to a husband rather than a family of origin. Sacrificing their family surname was a marker of this natural transition. They invoked religious language to justify the transition. They spoke of "becoming one" with a male spouse, which is a reference to biblical passages such as Matthew 19:4–6 that describe God's intent for a man to leave his father and mother and unite with his wife to become "one flesh".

"I think that it is important for me to commit to my future husband by taking his last name. I also believe it's important for the whole family, husband, wife, and children to have the same name. Husband and wife are called to become 'one flesh,' and for me the last name is a part of that." (female student)

 "I want to use the last name of my spouse because I want to show that I'm committed to him and it represents the oneness of our marriage." (female student)

Tradition was a third recurring theme in students' explanations. Tradition is a common reason given for marital name change by women [22]. Religion seems to accentuate the importance of tradition in this regard [15]. The appeal of tradition was evident within the highly religious sample of students in our study. More than 30% alluded to tradition as guiding their plans for traditional marital name change. Students referenced cultural norms and familial expectations. Several poignant examples were:

"I plan to change my last name when I get married to that of my spouse's because it is tradition. My mother did it, my grandmother did it, my great-grandmother did it, *etc.* I think it is nice and traditional for the woman to change her name." (female student)

"It has always been that way. It is just tradition that she will take my name and respect the tradition of my family. Why alter the tradition?" (male student)

"I just feel that it is the 'norm' in our society and I would feel weird if I didn't take his last name when we got married." (female student)

Even in the absence of explicitly religious language, these students reveal the religious influences at work in their decision-making. They plan to marry a spouse of the opposite sex and follow traditional naming practices. They justify their support for marital name change on the basis of gender roles, identity, and tradition. Their responses hint at shared moral boundaries governing gender, marriage, and family. Moral boundaries defining appropriate patterns of behavior are a hallmark of American Evangelicals [41]. In line with their Evangelical upbringing and present location in Evangelical colleges, students articulate values, morals, and ideals of their subculture.

### **7. Conclusions**

Marriage and family are contested terrain in American society. The purpose of this study was to examine marital naming plans among students in Evangelical colleges. Changing views on marital surnames in heterosexual marriages has been a topic of research for several decades. In an extension of previous research, we focus more directly on religion and religious context. Our survey of 199 introductory sociology students at four Evangelical colleges across the United States makes clear the ubiquity of heterosexual marriage in the Evangelical subculture. Only one out of 199 students did not plan to marry. Marriage is an anticipated rite of passage for these students. Choosing to attend an Evangelical college may be an intentional step toward this rite. Finding a spouse may not be in the promotional materials, but it is a selling point for these colleges.

When the wedding does arrive for these students, they plan to follow the naming pattern of their parents and larger society. Male and female respondents in almost equal percentages planned to follow a traditional naming pattern when married. Surprisingly, based on limited comparison data, students in our sample did not appear more traditional in their marital naming plans than students at other non-religious colleges. Like the desire to marry, husband and wife sharing the man's last name is a tradition that continues to appeal to many students. Our interest was whether religious beliefs and behaviors make patrilineal descent more appealing to students at Evangelical colleges. We tested a wider range of religious influences than considered in previous research. Our findings singled out two religion variables as important to students' marital naming plans: prayer and view of the Bible.

Students in our sample who reported praying more frequently indicated an intention to follow the marital naming custom for their gender. This is an interesting finding that contrasts the conflicting results over religious service attendance in previous studies. Unlike church attendance, prayer is a measure of private religiosity. Prayer connects an individual to God in a personal way. The influence of prayer then rests heavily on how individuals conceive of God and the will of God. For Evangelicals, the conception of God is as an authoritative presence who acts to ensure that humanity follows divine ideals [42]. These ideals include a gendered order to reality. Evangelicals who pray regularly might be understood as being most invested in upholding the created order. Consequently, it may be that young men in our sample are praying for someone to take their last name, while young women pray for a name to take. Our survey data do not tell us the content of prayers. Given the nearly unanimous desire to marry among these highly religious students, it is not an unreasonable speculation.

A belief in the gendered order of reality likewise helps explain the significance of Biblical literalism in predicting marital naming plans for students at Evangelical colleges. Students in our study who stress the inerrancy of Christian scripture envision a future for themselves in which they take a traditional marital surname. The relationship between view of the Bible and traditional marital surnames is significant and strong, even when controlling for other influences. Our findings in this regard are a logical extension of previous research that shows literal views of the Bible associated with more traditional views of gender [25,43,44] and more traditional attitudes toward marital namechange [11]. Our findings on Biblical literalism are important for another reason. They demonstrate that differing beliefs about the Bible have implications even within a single religious tradition. Evangelicals as a religious category have a strong emphasis on the authority of Scripture, but the views of Evangelical individuals are not uniform. Beliefs about the Bible, in particular, are a salient feature shaping individual's opinions within and across religious contexts.

What is perhaps most notable in our findings is the absence of religious language in students' explanations of their marital naming plans. Most of these highly religious young adults did not directly attribute their preferred marital surname to God, the Bible, or their churches. Instead, we see evidence of an invisible religious influence guiding these students' martial name-change decisions. Regnerus developed a typology of religious influence to account for the ways that religion impacts the sexual beliefs and practices of American teens [45]. One of the most common forms of religious influence in the lives of teens is what he called "invisible religion" ([45], pp. 194–96). In contrast to "intentional religion" in which individuals recognize and articulate the ways that religion directs behavior, invisible religion refers to the way that religion can guide human thought and action unbeknownst to an individual. A teen may explain her decision to postpone having sex by mentioning concern for her reputation or a desire not to disappoint her parents. These are not explicitly religious reasons, but her behavior does align with religious norms. The comments we heard from students regarding their marital naming plans similarly reflect religious origins. Concepts of gender roles (*i.e.*, headship) and identity (*i.e.*, becoming one) were prominent in students' explanations. Young men expected to keep their surname at marriage because "I am the man". Young women willingly planned to take on a male surname as a sign of "becoming one". Tradition is appealing to these students. In all of these ways, the explanations speak to an understanding of gender and family promoted within the Evangelical subculture.

Regnerus pointed to religious families as a prominent source of invisible religious influence [45]. Here again, our findings concur. Students in our study came, by and large, from religious families. Their choice of an Evangelical college speaks to their religious heritage. Not coincidentally, the family of origin for these students conformed to traditional patterns. Mother and father were married and shared the same last name. This model of family life molds students' expectations. Mother's surname was significantly related to plans to follow a traditional naming practice, even after controlling for other religious and demographic variables. As modeled by their parents, these Evangelical college students expect to marry and they expect to share the husband's last name. The belief that one's family expects such a naming pattern increases the propensity for a student to say that this is their intent. It is so taken for granted that some students don't perceive a choice. "I've never been taught that there was another option", wrote one female respondent. Through religious families, religious ideals for marriage and family get passed on. These ideals are formative. They do not come exclusively from prior religious socialization however. Present religious practices and beliefs are formative for individuals as well. Taking past socialization into account, private prayer and view of the Bible still stand out as significant predictors of marital naming plans in our multivariate analysis. Religious socialization past and present elevates the influence of religion, even if invisible.

We acknowledge limitations in our study. Our sample is not representative of all college students or all students in Evangelical colleges. More detailed analysis across religious traditions represents a fruitful direction for additional research. Likewise, we acknowledge the limitation of cross-sectional data for testing implied causal relationships. We argued that religion is a source of worldview that helps shape how individuals understand gender and marriage, but we recognize that these realms of social life and human identity are likely mutually reinforcing. Disentangling the causal order of religion, gender, and marriage is not something we can accomplish in this study. We hope future research will. Given the focus on heterosexual marriage in our study and others, another useful extension of this research will be to explore surname choices within same-sex marriages.

Although we cannot generalize from our convenience sample to all CCCU schools or all CCCU students, our findings suggest that traditional views of marital naming are common on these campuses. Like family, peer groups are powerful agents of socialization. The small, residential colleges in our study pride themselves on fostering tight-knit campus communities. Students become part of a campus culture with shared beliefs, values, and practices. These shared ideals guide individuals, as seen in regard to student religiosity. At a time in life when many young adults distance themselves from religion [36], students in our sample remained devout. We believe similar socialization occurs for gender and marriage on these campuses. By surrounding an individual with others who embrace more traditional views of gender and marriage, Evangelical colleges presumably lead students to conform to traditional gender scripts. The vast majority of students surveyed plan to follow traditional marital naming customs. Consequently, it may be hard for the minority who desire an alternative marital name to find a like-minded mate among their classmates. Future research should explore peer expectations as well as differences between religious colleges and non-religious colleges in the choice of marital surnames. Our guess is that many progressive-minded students in Evangelical colleges end up adhering to custom. In this way, Evangelical colleges work in tandem with Evangelical churches to perpetuate traditional ideals of marriage and family.

### **Acknowledgments**

The authors thank Charles Tolbert, Cassidy Cooper, Jane Scott, F. Carson Mencken, Jeremy Uecker, Wade Rowatt, and faculty collaborators at the four colleges for contributions to this research.

### **Author Contributions**

Kevin Dougherty designed the study, coordinated data collection, and conducted statistical analysis. Melanie Hulbert participated in data collection and conducted qualitative analysis. Ashley Palmer created the survey instrument. All three authors contributed to writing the manuscript.

### **Abbreviations**

CCCU: Council for Christian Colleges and Universities.

### **Conflicts of Interest**

The authors declare no conflict of interest.

### **References**


### **The Optimal Level of Strictness and Congregational Growth**

### **Todd W. Ferguson**

**Abstract:** Beginning with Kelley's and Iannaccone's foundational studies, scholars have examined how strictness impacts congregational outcomes. This paper seeks to further develop the strict church thesis by examining Iannaccone's concept of "optimal level of strictness", an idea that there are limits to strictness. Using Stark and Finke's theoretical framework of religious niches and data from the 2005 Baylor Religion Survey and the 2000 Faith Communities Today survey, I find that only prohibitions that are in line with a congregation's religious niche have an impact on growth. To be beneficial, prohibitions must match the pool of potential members' preferences.

Reprinted from *Religions*. Cite as: Ferguson, T.W. The Optimal Level of Strictness and Congregational Growth. *Religions* **2014**, *5*, 703–719.

### **1. Introduction**

In the 1960s and early 1970s Liberal and Moderate Protestants were struggling to find a reason why their churches were declining in numbers. For much of the history of the United States, Presbyterians, Methodists, Episcopalians, and Congregationalists were at the center of American religious life. Yet, in the mid-twentieth century, these denominations found their churches losing members and losing prominence. At the same time, churches in conservative denominations, such as the Church of the Nazarene, the Southern Baptist Convention, and the Assemblies of God, were growing both in number and in their percentages of the total population. To help understand this change, Dean Kelley, a researcher with the National Council of Churches, proposed the strict church thesis. Going against the predominant assumption that modern individuals would gravitate toward an open-minded and tolerant religion, Kelley put forward the opposite. Strict congregations—those with more rules and more exclusive claims—are stronger because they are better than the more lenient churches at offering meaning for their members [1].

Laurence Iannaccone [2] furthered Kelley's theory by incorporating economic mechanisms to Kelley's cultural argument. He defines strictness as "the degree to which a group limits and thereby increases the *cost* of nongroup activities" ([2], p. 1182, emphasis in original). Religious groups vary in the ways they create strictness, and this diversity can be seen in how Seventh Day Adventists avoid eating meat, Orthodox Jews wear side curls and yarmulkes, and Jehovah's Witnesses reject blood transfusions. Iannaccone suggested the reason why strict congregations are strong is because they reduce the problem of free-riders. Religion can be viewed as a commodity that is produced with others in community. The satisfaction an individual derives from religious practice depends on the quality that others produce. Free-riders are those who do not add anything to the collective religious product but nevertheless reap its benefits. Free-riders, therefore, lower the benefits-per-individual in the congregation. Strict congregations reduce free-riders because they raise the cost of participation. These churches "penalize or prohibit *alternative* activities that compete for members' resources" ([2], p. 1187, emphasis in original). These prohibitions serve as entry fees for participation and screen out members who might not fully add to the collective religious product. Consequently, strict congregations have fewer free-riders, have higher benefits-per-individual, and experience more congregational strength.

Although Kelley's [1] original book was titled "Why Conservative Churches Are Growing", neither Kelley nor Iannaccone [2] focused on congregational growth. Growth was only a by-product of strength. Yet further research has shown that strictness and congregational growth are indeed linked. Iannaccone, Olson, and Stark [3] use the theoretical framework of resource mobilization to understand the connection between strictness and growth. Religious congregations can grow only if they have a surplus of resources, specifically time and money. Strict congregations are better able to have surpluses of time because they restrict their members from engaging in alternative activities [2]. These members are more likely to focus on congregational activities because they have few other options. Similarly, these congregations often limit where members can spend money, and so there is more likely to be a surplus of financial resources. Strict congregations are better able to accumulate surpluses of time and money and therefore are more likely to grow. Thomas and Olson [4] confirm this finding that, even when accounting for fertility, congregational strictness significantly and positively affects growth.

Yet are there circumstances where strictness and congregational outcomes are not linked together? Kelley proposed that there are no limits to strictness; he states, "strong organizations are strict…the stricter the stronger" ([1], p. 95). Yet Iannaccone disagrees. Strictness "displays diminishing returns" ([2], p. 1201). There is a point in which an increased level of strictness will not see an accompanying level of either strength or growth. This is the "optimal level of strictness", which is based upon the characteristics of the congregants ([2], p. 1202). Going beyond the optimal level of strictness means that religious groups may build strictness around behaviors or attitudes that do not match the congregants' preferences. Congregants perceive these prohibitions as excessive, which inhibits growth. Therefore, strictness must be at its optimal level in order to affect congregational growth. This paper seeks to further the strict church research by examining the optimal level of strictness for congregations. I use the theoretical framework of niches to understand when the diminishing returns of strictness occur.

### **2. Religious Niches**

The optimal level of strictness corresponds to the preferences of the individuals within congregations. Individual members are the most important resource for religious organizations [5–8],<sup>1</sup> and they vary in their demand for religious goods and services, or their preferences [11,12]. Some prefer very strict religious organizations, while others lean toward more liberal congregations. Thus, individuals may be grouped into theoretical categories, or *religious niches*, based on their preferences for strictness. A religious niche is the "market segment of potential adherents sharing

<sup>1</sup> Individual members are the constituent resource for religious organizations because they not only are the suppliers of time and money that influence congregational growth [3], but they also impact ritual density, an important factor in sustaining religious communities [9,10].

particular religious preferences (needs, tastes, and expectations)" ([12], p. 195).2 Figure 1 shows Stark and Finke's [12] conceptualization of religious niches based on individual preferences for strictness. A major assumption is that these niches are stable over time because the demand for various levels of religious strictness is constant within the population ([14], p. 5). As shown in Figure 1, this assumed distribution of preferences may be thought of as a normal curve, with most of the population desiring some level of strictness that is neither too lenient nor too severe. This does not imply that *individual* preferences are nonmalleable. Individual preferences are culturally shaped and therefore, changeable. However, at the aggregate, preferences are assumed to follow a normal distribution.

Congregations and their denominations form around these niche categories to serve the religious demand [7]. Thus, they compete for resources, *i.e.*, members [6,8,13,15]. It would seem to follow that if a congregation's level of strictness does not match the individuals' preferences within a given niche, the congregation would necessarily recruit members from another niche. Yet the nature of congregations complicates this view. Congregations are more nuanced than aggregations of individuals that compete for potential resources [16]. They are "instantiations of larger institutions" ([5], p. 207). They are carriers of denominational culture, which shapes a congregation's level of strictness. Denominations are associated with certain levels of strictness, even though they exhibit considerable amount of internal diversity within themselves [12,15,17].

These denominational assumptions of strictness are, in turn, ascribed to the congregations. Even non-denominational congregations that do not formally connect with a group are still subject to prevailing expectations associated with larger religious traditions, most notably Conservative Protestantism. Thus, congregations are not entirely free to respond to the religious marketplace's demands from individuals. Many are constrained by their denominational culture and its assumed level of strictness.

Religious traditions are a useful way to represent niches within the American landscape. These are broader groupings of denominations that have similar beliefs, practices, and histories [18–20]. Their constituents have similar preferences in matters of strictness [12]. Though the division is contested, I follow Roof and McKinney [21] to categorize religious groups into eight religious traditions in the United States: Catholics, Jews, Conservative Protestants, Moderate Protestants, Liberal Protestants, Black Protestants, Other religious groups, and "Nones" (those without religious affiliation) but see [18,19]. While religious traditions are not entirely coterminous with niches in the American landscape, this categorization of religious groups as representing niches is useful and easily accessible.

<sup>2</sup> This is a slightly different understanding of niche than Popielarz and Neal ([13], p. 68), which they say is "the set of environmental states in which [a species] needs to survive." In this present study, niches are built around individual preferences, following Stark and Finke [12]. The definitions are most similar at the organizational level, since members, which are organized around preferences, are considered the resources necessary for congregational survival.

The first two traditions are eponymous, as these groups consist of all Catholic and Jewish groups. Protestants fall into four groups. The Conservative Protestant tradition is comprised of groups that are more sect-like, which emphasize Biblical authority, missionary activity, and individual conversion [20,22,23]. The Liberal Protestant tradition focuses more on "an accommodating stance toward modernity, a proactive view on issues of social and economic justice, and pluralism in their tolerance of varied individual beliefs" ([19], pp. 293–94). Moderate Protestants fall in between these two, as these denominations have congregations that lean toward both of these categories [21]. Black Protestants are the groups that have been shaped by the African-American experience [24]. "Other religious groups" consist of Jehovah's Witnesses, Latter-Day Saints (Mormons), Buddhists, Hindus, and Muslims. Religious "nones" are a growing category which scholars see as an emerging, distinct group [25–28].

### **3. Religious Niches and Congregational Growth**

In order for a congregation to grow, it must gain new members. Potential members are most likely to come from within the congregation's religious niche and not from another niche. This is because people are less likely to move out of their current religious tradition and join a congregation in another [11]. Most people join religious congregations that are very similar to their previous one or match their desired level of preference [12]. There is a "homophily of preferences", and so individuals retain as much religious capital as they can by choosing a new congregation ([12], p. 195). A move to another religious niche would cost too much religious capital, and so individuals usually stay within congregations that are similar in strictness. As a result, the pool from which congregations draw potential members is most likely the congregation's very own religious niche.

Therefore, it is possible that a congregation's level of strictness could be suboptimal if it is disconnected from its own religious niche. That is, there could be a mismatch between potential members' preferences for strictness and the congregation's own practices. When a congregation's level of strictness is outside the preferences of potential members, then growth is less likely to occur because the congregation is less likely to be attractive within its religious niche. Based on this, I offer the following hypothesis: *Strictness increases the likelihood of congregational growth when it is in line with the preferences of the individuals within the congregation's niche.* 

### **4. Testing the Hypothesis**

The relationship between strictness and congregational growth is contingent upon the preferences of those within a religious niche. In order to test this hypothesis, I follow a two-step process. First, I ascertain how people within religious traditions vary in their preferences on an issue of strictness. In essence, I am mapping the contours of religious niches for a specific issue of strictness. Second, I test to see if these preferences affect the relationship between strictness and congregational growth.

For this present study, I examine three religious traditions within American Protestantism (Conservative, Moderate, and Liberal Protestants) and their preferences regarding two historically important issues: alcohol and premarital sex. Protestants do not hold a single view on these two subjects. Their preferences vary, which create distinct religious niches indicating different levels of desired strictness. This offers a useful basis to test the relationship between religious niches and strictness.

### **5. Religious Niches for Alcohol and Premarital Sex**

Following the end of the Civil War, many Protestants, especially Methodists and Baptists, turned their attentions to reforming the misuse of alcohol [29]. They created temperance movements across the country and succeeded in pushing for the Eighteenth Amendment to the Constitution which prohibited alcohol in the United States. Although the Twenty-first Amendment overturned the Eighteenth, the trajectory initiated by the temperance movement still remains in much of American Protestant life. Protestant religious beliefs about alcohol are changing, but there continues to be a wide variation within these beliefs. Some Protestants consider all alcohol use to be wrong, although others do not think drinking alcohol is an issue [30].

In order to examine strictness preferences within religious niches, I use the 2005 Baylor Religion Survey (BRS), which is a useful data set to determine how individuals view alcohol. This survey is a national random sample of 1721 individuals in the contiguous United States which the Gallup Organization administered using a mixed-mode method in October and November 2005. Bader, Mencken, and Froese [31] outline the full methodological information. Others have used this survey successfully to understand how various religious groups differ in their views of moral issues [32]. For this study, I examine whether individuals within a religious tradition have significantly different preferences about an issue when compared with those without a religious

tradition. This allows me to estimate the boundaries of the religious niche in terms of a single issue of strictness. Others have used a similar method with national surveys of individuals to outline religious niches [8].

Figure 2 shows how Americans in three Protestant traditions—Liberal, Moderate, and Conservative <sup>3</sup> —compare with Americans with no religious affiliation when answering the following question on the BRS: "How do you feel about the consumption of alcohol?" Individuals could respond (4) Always wrong, (3) Almost always wrong, (2) Only wrong sometimes, and (1) Not wrong at all. Following conventional niche measurements, the mean responses plus or minus 0.75 standard deviation are displayed [6,8]. Although the responses for each religious tradition show both diversity within and niche overlap between traditions, a post-hoc analysis of an ANOVA test using Scheffé's method reveals that Conservative and Moderate Protestants hold significantly different views than individuals without religious affiliation on alcohol consumption, thus indicating different religious niches. Stated differently, the potential members for Conservative and Moderate Protestant churches are located within religious niches that desire some level of strictness surrounding alcohol. Liberal Protestants, however, are not significantly different than those without religious affiliation, indicating that those within this religious tradition do not prefer strictness on this issue. They are located within a religious niche for which alcohol is not an issue.

Premarital sex is also another relevant issue for American Protestants. Traditional Christian teaching has placed sexual activity within the confines of heterosexual marriage. However these boundaries have been challenged beginning in the late 1960s with the sexual revolution. Figure 3 shows how individuals without religious affiliation and Protestants from Liberal, Moderate, and Conservative denominations answered the following question from the 2005 Baylor Religion Survey: "How do you feel about sexual relations before marriage?"

Figure 3 shows the ordering of the groups' beliefs about premarital sex is the same as alcohol. The average responses for people without religious affiliation and Liberal Protestants are lower than Moderate Protestants, who in turn, are lower than Conservative Protestants. The difference between beliefs about alcohol and premarital sex, however, is that all three religious groups—Liberal, Moderate, and Conservative Protestants—are significantly more likely to classify premarital sex as wrong, than those without religious affiliation. Therefore, each tradition is within a religious niche that prefers some strictness surrounding the issue of premarital sex.

To summarize the first step for testing the hypothesis, individuals within all three Protestant traditions have preferences for strictness surrounding premarital sex when compared to those

<sup>3</sup> I created religious tradition categories to match the religious traditions in the Faith Communities Today (FACT) survey, which will be used in the analysis below [33]. For BRS data, Liberal Protestants (*n* = 173) are Congregational, Episcopal/Anglican, Presbyterian, Unitarian Universalists, and United Church of Christ. Moderate Protestants (*n* = 273) are: American Baptists, Disciples of Christ, Evangelical Lutheran Church of America, Mennonite, United Methodist, and Reformed Church of America/Dutch Reformed. Conservative Protestants (*n* = 393) are Assemblies of God, Southern Baptist, Christian Reformed, Church of Christ, Church of the Nazarene, Seventh-day Adventist, and Non-denominational Christian. There are 192 respondents without religious affiliation. I had to omit Black Protestants from analysis because FACT data did not contain the necessary variables for this religious tradition.

without a religious affiliation. They are within religious niches that prefer strictness for premarital sex. On the other hand, only Moderate and Conservative Protestants have distinct preferences for strictness surrounding the issue of alcohol. Thus, those within the religious niche for Liberal Protestants do not desire strictness for this issue.

**Figure 2.** Individual Views on the Consumption of Alcohol (Means of responses ± 0.75 standard deviation). All differences are significant at the 0.05 level except *No Affiliation—Liberal Protestants* and *Liberal Protestants—Moderate Protestants*.

Source: Baylor Religion Survey, 2005.

**Figure 3.** Individual views on Premarital Sex (Means of responses ± 0.75 standard deviation). All differences are significant at the 0.05 level.

Source: Baylor Religion Survey, 2005.

### **6. Testing Optimal Strictness**

The next step in testing my hypothesis is to see if there is an optimal level of strictness for congregations and growth. In other words, my goal is to see if the relationship between strictness and growth continues in circumstances when the religious preferences of a congregation's potential members (its religious niche) do not line up with congregational practices.

To do this, I use the Faith Communities Today (FACT) survey from 2000, which I obtained from the Association of Religion Data Archives [34]. Coordinated by the Hartford Institute for Religion Research, this survey is the largest study ever conducted on congregations in the United States [33]. FACT data represent 41 denominations and faith groups. Each faith group's survey included core questions on six areas of congregational life and structure: worship, location, programs, leadership, participants, and finances. An informed respondent, usually the senior clergy person, filled out the survey. The survey response rate for the denominations averaged just over 50 percent. A total of 14,301 congregations completed surveys.

FACT divides religious congregations into six categories similar to Roof and McKinney's [21] categories: Liberal Protestant, Moderate Protestant, Conservative Protestant, Historically Black, Catholic/Orthodox and Other. Because of the limited nature of the congregational discourses used to create measures of strictness (discussed below), I remove the Catholic/Orthodox and Other categories from the sample. Furthermore, the survey instrument for Historically Black Protestant congregations unfortunately did not contain many of this study's variables. Therefore, I restrict my analysis to Liberal, Moderate, and Conservative Protestants. Liberal Protestants are Episcopal Church USA, Presbyterian Church USA, Unitarian-Universalist, and the United Church of Christ. There are 2565 Liberal Protestant responses. Moderate Protestants are American Baptist Churches, Disciples of Christ, Evangelical Lutheran Church in American, Mennonite, Reformed Church in America, and the United Methodist Church. There are 3263 Moderate Protestant congregations in the study. Conservative Protestants are Assemblies of God, Christian Reformed Church, Church of the Nazarene, Churches of Christ, Independent Christian Churches (Instrumental), Mega-churches, Nondenominational Protestant, Seventh-day Adventist, and the Southern Baptist Convention. There are 3610 Conservative Protestant responses. The FACT 2000 public data file from the Association of Data Religion Archives did not include a denomination variable. Therefore, I am forced to use categories described above.

FACT data are useful to test my hypothesis. First, these data are from a very broad sample. While not all faith groups in the U.S. participated, most of the largest denominations are included. Because of this breadth, FACT represents 80% of all U.S. congregations [33]. Second, FACT is a survey of some depth. Each congregation reported on church growth, how much they address social issues, how active they are at outreach, and the demographics of the congregants. In addition to the congregational questions, FACT includes United States Census data at the ZIP code level for 1980, 1990, and 2000. This allows me to control for the surrounding ecological influences for each congregation that might influence growth, such as a growing community population.

Third, this analysis requires a high number of responses from Liberal Protestant congregations who have either anti-alcohol or anti-sex discourses, which is not possible with the other national congregational surveys, such as the National Congregations Study and the U.S. Congregational Life Survey.

### *6.1. Dependent Variable: Congregational Growth*

To measure congregational growth, FACT asked each congregation, "Since 1995, has the number of regularly participating adults: Decreased 10% or more; Decreased 5% to 9%; Stayed about the same (plus or minus 4%); Increased 5% to 9%; Increased 10% or more?" I create a binary church growth variable for (1) congregational growth of 5% or more and (0) congregational stability or decline. The appendix shows the descriptive statistics for each variable.

### *6.2. Independent Variable: Congregational Discourse on Alcohol Use and Premarital Sex*

Congregational strictness is measured by the level of discourse surrounding alcohol use and premarital sex within a church. The FACT survey asked, "How much does your congregation, in its worship and education, emphasize the following home and personal practices?" Practices included personal prayer, family devotions, fasting, observing a special diet, abstaining from alcohol, observing a weekly holy day, displaying icons, and abstaining from premarital sex. I focus on abstaining from alcohol and premarital sex because of their relevance to American Protestants. Each congregation could respond "Not at all", "A little", "Some", Quite a bit", or "A great deal". Tables 1 and 2 show how congregations in each Protestant tradition responded. The distributions of anti-alcohol and premarital sex discourses within congregations mirror the individual responses to these prohibitions. Conservative Protestant churches are the most likely to actively emphasize antialcohol and premarital sex messages. Liberal Protestant churches are the least likely.


**Table 1.** Congregational Discourse on Abstaining from Alcohol.

Source: Faith Communities Today (2000). Totals do not equal exactly 100% due to rounding.

In the analysis that follows, I create binary, *prohibition discourse* variables for anti-alcohol and anti-premarital sex to measure congregational emphases on these two behavior issues. Coding is (1) Congregations that emphasize the topic "Quite a bit" or "A great deal" and (0) congregations that emphasize the topic at the other levels.

**96**


**Table 2.** Congregational Discourse on Abstaining from Premarital Sex.

Source: Faith Communities Today (2000). Totals do not equal exactly 100% due to rounding.

### *6.3. Congregational Control Variables*

I control for other congregational factors that might contribute to church growth. The age of a church is held constant by using the year founded. Congregation size is measured by the natural logarithm of the number of adults (18 years and older) who regularly participate in religious life at the congregation, whether or not they are members. I also control for the demographics of a church by holding constant the percentage of younger adults (under 35), older adults (over 60), and females participating in church life. These three variables range from 1 to 7 with the responses being (1) None 0%, (2) Hardly any 1%–10%, (3) Few 11%–20%, (4) Some 21%–40%, (5) Many 41%–60%, (6) Most 61%–80%, and (7) All or nearly all 81%–100%. Because the responses 1 through 7 are not meaningful as a numeric scale, I use the midpoints of each response (0, 5, 15, 30, 50, 70, and 90) to create interval variables.

I also control for congregational outreach activities. The FACT survey asked: "In addition to the outreach activities of your denomination, did your congregation do any of the following during the past 12 months to reach out to new or inactive participants, or to make your congregation better known in your community?" I control for three types of marketing approaches: newspaper ads, radio and television ads, and direct mail promotions. Each of these is a binary variable: (1) Yes, done in the last 12 months and (0) No.

### *6.4. Community Control Variables*

In addition to the internal factors within a congregation, I also control for a congregation's surrounding community. I control for ZIP-code population in 2000 (natural logarithmic transformed) and the percentage change in ZIP-code population from 1990 to 2000. This variable is a discrete, 1 percent interval measure of the percentage change (e.g., −12% or 5%). The upper and lower ranges are capped off at "−20% or lower" and "30% or higher" (coded as −20 and 30, respectively). Finally, region of the country is held constant by a series of binary variables, with South as the comparison group.

### **7. Method**

Binary logistic regression is the most appropriate method because the dependent variable of church growth is dichotomous. Because of the high correlation between anti-alcohol and anti-premarital sex discourses (*r* = 0.52), I separate these independent variables in the models. I estimate six models, separating the three Protestant traditions by how anti-alcohol or anti-premarital sex discourses affect church growth.<sup>4</sup> I expect significant positive estimates for the relationship between growth and a congregation's strictness on an issue for religious traditions whose pool of potential members have distinct strictness preferences (e.g., the issue of alcohol for Conservative Protestants or premarital sex for Liberal Protestants). On the other hand, I expect non-significant estimates for the relationship between congregational growth and an issue of strictness when a religious tradition's potential members do not have preferences of strictness that are distinct from the unaffiliated.

### **8. Results**

Table 3 shows the results from six binary logistic regressions predicting a congregation growing 5% or more. As hypothesized, only the congregational discourses about strictness that match its niche's preferences are the ones that are associated with increased odds of being a growing congregation, while the one that does not match does not affect the likelihood of growth.

Models 1, 3, and 5 show anti-alcohol discourse affects the probability of congregational growth in both Moderate and Conservative Protestant churches, but *not* in Liberal Protestants congregations. Liberal churches that actively emphasize anti-alcohol messages are no more likely to grow than those who do not maintain this prohibition. By comparison, Conservative churches with high levels of anti-alcohol discourse are 73% more likely to grow, and Moderates with the same level are 55% more likely.

Other congregational characteristics also affect church growth. Churches founded more recently and larger congregations (*i.e.*, those with more regularly participating adults) are more likely to grow. Churches from all traditions with higher percentages of older adults have lower odds of growth, as do Liberal and Moderate Protestant congregations with more females. Advertising through radio, television, and mail outs only affects growth for Conservative congregations. Liberal and Conservative churches in areas with greater ZIP populations have lower odds of growth, but a growing ZIP code population increases the odds for all three traditions. Liberal and Conservative churches in the Midwest are less likely to grow compared to Southern Liberals and Conservatives.

Unlike anti-alcohol discourse, which did not affect congregational growth in all traditions, anti-premarital sex discourse does. Models 2, 4, and 6 in Table 3 show Liberal, Moderate, and Conservative Protestant congregations that have high levels of anti-premarital sex discourse are 55%, 34%, and 54% more likely to grow, respectively. Like the other models, newer churches and larger churches are more likely to be growing, while the percentage of older adults lowers the odds of growth for all three traditions. The percentage female only lowers the odds of growth in Liberal Protestant congregations.

<sup>4</sup> I also estimated the models using an OLS regression. I used 0.10, 0.07, 0.00, −0.07, and −0.10 for the measurement of congregational growth. The results were the same as the binary logistic regression. The tables are available upon request.

The gender ratio has no effect in Moderate and Conservative churches in these models. Advertising through radio, television, and the mail increases the odds of growth in Conservative congregations, but has no effect for Moderates and Liberals. Community variables also affect the odds of congregational growth. Liberal and Conservative Protestant churches in ZIP codes with large populations have lower odds of growth. ZIP code population does not make a difference for Moderate congregations. However, the growth in congregational ZIP codes does affect the odds for all three traditions. For each percentage increase in ZIP code population change, the odds that a church is growing increase by either 1% or 2%. Finally, congregations in all traditions are less likely to grow if they are located in the Midwest as compared to the South.


**Table 3.** Binary Logistic Regressions Predicting Congregations Growing 5% or More (Odds Ratios Shown).

Note: \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001. Source: Faith Communities Today (2000).

### **9. Discussion and Conclusions**

This study seeks to better understand the relationship between congregational strictness and growth. I have done so by showing evidence for the idea that strictness must be in line with the congregation's religious niche, its source of potential members. As others have found before, *strictness still matters* [1,2,4,12]. The boundaries that churches create by establishing rules raise the costs of being a member. These increased costs allow the church to generate higher levels of resources, which in turn increase the chances for growth [3]. Moreover, strictness is not a factor for only Conservative Protestants; it is important across all three Protestant traditions.

More importantly, this study modifies the classic strict church thesis. It helps shed light on what Iannaccone [2] meant by an optimal level of strictness. Strictness is not *always* associated with congregational growth. That is, a congregation cannot prohibit any behavior and expect this increase in strictness to increase the likelihood of growth. Instead, growth via strictness is contingent upon potential members' preferences for strictness on an issue. This is why the Liberal Protestant congregations that have high levels of anti-alcohol discourse are not more likely to grow.

These churches, by definition, are stricter churches than their fellow Liberal congregations, but this strictness does not affect growth. Those in their religious niche, the greatest source of potential members, do not want a church with rules about drinking. Thus, the costs incurred by prohibiting alcohol are too high for Liberal Protestants. Potential members are unwilling to pay this price of membership because they do not prefer a level of strictness on this issue.

The prohibition of alcohol does affect growth for Moderate and Conservative congregations, though, because the costs incurred are much lower. In fact, this restriction might actually be a benefit because it lines up with the preferences of their religious niches. Therefore, the prohibition of alcohol in these congregations may be viewed as both a cost and a benefit. It is a cost that restricts alternative behavior and frees up other resources conducive for growth, but it also creates a favorable cost-benefit ratio for its potential members by lining up with their niche preferences.

These findings also support the idea that there is a "spiritual marketplace" in American Protestant Christianity [35,36]. Religious niches are essentially constructed around consumer preferences, and congregations market and supply the religious goods to these niches. If a congregation is out of step with its niche (*i.e.*, its practices and discourses used to construct its level of strictness do not match the demands of the religious marketplace), then it will not attract new members from the niche. For this reason, Liberal Protestant churches who emphasized not drinking alcohol were not likely to be growing congregations.

This study has some limitations. FACT is not a random sample of congregations throughout the United States. Instead, it is a very broad survey of participating faith groups. Positively, this means that the number of congregations and denominations in this study is quite high [37]. Negatively, it means that, while strongly suggestive of how strictness works within congregations, these findings are not nationally representative of all American Protestant churches.

The main drawback of the 2000 FACT survey is its limited scope. The survey asked about only two prohibitive discourses that are salient for Protestant Christians in the United States: alcohol and premarital sex. These are by no means the only relevant issues for Protestants, and Protestantism is by no means the only religious tradition. This study could be strengthened by examining congregations in other countries and looking at other issues, such as theological beliefs, economics, race, gender, or sexuality. It could also be strengthened to see where the line for optimal strictness is in other congregationally-based religious traditions, such as Catholicism, Islam, and Judaism. Furthermore, these findings open up the possibilities to examine how optimal strictness, as delineated by religious niches, impacts congregational *strength*, the main focus of Kelley [1] and Iannaccone [2].

Future research in the sociology of religion needs to account for the idea that strictness is a multifaceted concept. It does not uniformly affect other congregational outcomes. An organization's niche matters, and the population from which the congregation draws new members responds differently to numerous issues used to create congregational strictness. Scholars continuing to explore the strict church thesis must take this into account when they explore how a religious organization's rules and regulations impact congregational life.

This study advances the classic strict church thesis within the sociology of religion by illuminating what optimal strictness is for congregational growth. Congregational prohibitions do not automatically increase the likelihood of growth. Effective prohibitions are contingent and defined by the congregation's religious niche (*i.e.*, its potential members). By lining up its congregational practices with its religious niche's preferred level of strictness, a congregation increases its chances for growth.

### **Acknowledgments**

The author would like to thank Kevin Dougherty, Paul Froese, Charles North, and Christopher Pieper for their valuable assistance.

### **Conflicts of Interest**

The author declares no conflict of interest.

#### **Appendix**


Descriptive Statistics for Variables.

Source: Faith Communities Today (2000).

### **References**



### **Neighbors Like Me? Religious Affiliation and Neighborhood Racial Preferences among Non-Hispanic Whites**

### **Stephen M. Merino**

**Abstract:** Research on racial residential segregation has paid little attention to the role that social institutions play in either isolating or integrating racial and ethnic groups in American communities. Scholars have argued that racial segregation within American religion may contribute to and consolidate racial division elsewhere in social life. However, no previous study has employed national survey data to examine the relationship between religious affiliation and the preferences people have about the racial and ethnic composition of their neighborhoods. Using data from the "Multi-Ethnic United States" module on the 2000 General Social Survey, this study finds that white evangelical Protestants have a significantly stronger preference for same-race neighbors than do Catholics, Jews, adherents of "other" faiths, and the unaffiliated. Group differences in preferences are largely accounted for by socio-demographic characteristics. Negative racial stereotyping and social isolation from minorities, both topics of interest in recent research on evangelical Protestants and race, fail to explain group differences in preferences.

Reprinted from *Religions*. Cite as: Merino, S.M. Neighbors Like Me? Religious Affiliation and Neighborhood Racial Preferences among Non-Hispanic Whites. *Religions* **2011**, *2*, 165–183.

### **1. Introduction**

Understanding and explaining racial residential segregation, which is thought to affect the range of opportunities available to minorities, remains an important task for researchers [1,2]. Explanations of racial residential segregation have generally focused on economic differences between racial and ethnic groups that influence residential options, [3-5], discriminatory practices in economic institutions and in the housing market that create and maintain segregation [1,6-8], and preferences people hold about the racial composition of neighborhoods in which they want to live [9-13]. Interestingly, little attention has been paid to the role that social institutions play in either exacerbating or ameliorating racial residential segregation. Given their important role in community and civic life, religious organizations are uniquely poised to influence intergroup relations.

Scholars have noted the potential for religious organizations to facilitate civic participation and community building [14-16]. There are indications that religious congregations sometimes play a role in supporting and facilitating racial integration in American communities [17,18]. However, religious congregations represent one of the most racially segregated institutions in the United States. The vast majority of congregations are essentially uniracial [19]. This deep racial divide has led some scholars to suggest that religion has the potential to consolidate racial division elsewhere in social life [20-22].

A recent study highlights the need for scholars to consider religion in research on racial residential segregation. Based on an analysis of county-level data from the 2000 U.S. Census and the 2000 Religious Congregations and Membership Study, Blanchard [23] reported that the number of evangelical Protestant congregations per 1,000 non-Hispanic whites was positively associated with levels of black-white residential segregation in both metropolitan and nonmetropolitan areas. This key finding was confirmed using measures of two different dimensions of residential segregation and shown to occur across all regions of the United States. Blanchard's 'closed community thesis' contends that while religious institutions sometimes foster bridging ties that link groups and promote social integration, white evangelical Protestant congregations' strong network closure and internal orientation prevent them from doing so.

Blanchard's study is an important call for more research on the role that religion and religious institutions play in facilitating or inhibiting social integration in communities. While Blanchard's study provides insight into how religious institutions may help integrate or isolate groups in a community, it does not directly examine the relationship between individuals' religious affiliation and their preferences about the racial and ethnic composition of their communities. While the roots of racial residential segregation are complex, there is strong evidence that neighborhood racial preferences are a significant contributing factor. Studies have consistently demonstrated that whites have a stronger preference for same-race neighbors than do minorities [8], and whites' avoidance of predominantly black or racially mixed neighborhoods is thought to uphold segregation [24-26]. While neighborhood preferences measured by surveys likely have an imperfect relationship with behavior, they are nonetheless useful in assessing individuals' willingness to live in integrated neighborhoods. Using data from the 2000 General Social Survey, this study will examine the relationship between religious affiliation and neighborhood racial-composition preferences among non-Hispanic white Americans. Furthermore, it will assess whether white evangelical Protestants' racial attitudes and social networks, both topics of study in recent research on race and religion, explain any observed distinctiveness in their neighborhood preferences.

### **2. Religion and Race**

While workplaces and public institutions have become increasingly integrated, religious congregations remain deeply segregated along racial and ethnic lines [27]. Emerson and Smith [20] contend that this segregation is due in part to the nature of the American religious market, which fosters competition, specialization, and individual choice. In addition, the authors contend, social psychological forces tend to push congregations toward internal similarity in order to facilitate the creation of symbolic boundaries and social solidarity. However, while segregation is the norm across religious traditions, there is some notable variation. Religious market share size plays a role [27]. The larger a religious tradition, the less racially diverse are its congregations. The lack of diversity is due to that fact that the more choice individuals have, the more exact they can be in realizing their preferences. Studies find multiracial congregations to be more common in Catholicism and non-Christian traditions than in Protestant denominations [27-29].

But does racial division within American religion have broader implications for society? Emerson and Smith [20] argue that it contributes to the racialization of America:

We claim that these patterns not only generate congregational segregation by race, but contribute to the overall fragmentation of American society, generate and sustain group biases, direct altruistic impulses to express themselves primarily within racially separate groups, segregate social networks and identities, contribute to the maintenance of socioeconomic inequality, and generally fragment and drown out religious prophetic voices calling for an end to racialization (p. 154).

Furthermore, the authors contend that the "stronger" the religion, the more it contributes to the racialization of society. Hence, their work focuses primarily on white evangelical Protestants' racial attitudes and beliefs [20,30].

Despite the important role that religious congregations play in local communities, little research has examined ways in which these social institutions may either challenge or maintain racial residential segregation. While Blanchard's study [23] provided evidence that the extent of racial residential segregation in a community is related to its congregational composition, it did not examine a factor thought to contribute to segregation: individuals' preferences about the racial composition of their neighborhoods [9-13]. How might religious affiliation affect neighborhood racial preferences? This study will focus on two ways in which religion could contribute to the racialization of American society. First, religious affiliation is thought to affect the racial composition of individuals' social networks, as well as the amount of interracial contact they experience. Second, religious traditions endow individuals with cultural tools that influence how they understand and interpret aspects of the social world, including race. By affecting individuals' opportunities for interracial contact and by influencing individuals' racial beliefs and attitudes, religion could shape the preferences people hold about the racial and ethnic composition of their neighborhoods.

### *2.1. Social Networks*

Emerson and Smith [20] contend that American religion contributes to the racialization of society by creating and reinforcing racially distinct social networks. By contributing to the separation of social life along racial lines, religion may reduce opportunities for developing intergroup social ties and bridging social capital. Furthermore, they contend that the "stronger" the religion, the greater the effect. The authors note that, during their interviews, they were "struck by how racially homogenous the social worlds of most evangelicals are" ([20], p. 80). Evangelical Protestant congregations tend to foster strong in-group ties that limit members' non-group activities and create dense intra-group social networks [31-34]. Evangelical Protestant congregations are also less involved in the provision of social services and offer fewer community outreach programs than congregations in other religious traditions [35-37]. The strong inward orientation of evangelical Protestant congregations could to lead to spatial and social isolation from minorities [20]. Conversely, individuals in traditions with more racially diverse congregations or greater civic involvement may have more opportunities for meaningful interracial contact.

The relationship between religious affiliation and social network diversity may have consequences for neighborhood racial-composition preferences. Spatial and social isolation from minorities lead to more negative racial attitudes and stronger in-group preferences. Oliver and Wong [38] find that individuals living in more racially homogeneous neighborhoods express more racial resentment than those living in more diverse neighborhoods. Significantly, even when

**109**

controlling for neighborhood preference, the negative effects of greater neighborhood racial homogeneity on out-group attitudes remain. In other words, self-selection into neighborhoods does not fully explain variations in out-group attitudes. Furthermore, individuals' prior experiences with interracial contact shape their future racial preferences. In short, racial isolation may breed future racial isolation, since people tend to choose what they have chosen or been assigned previously, a phenomenon known as the status-quo bias [39]. For example, those with prior interracial contact in schools and neighborhoods are more likely as adults to have more racially diverse general social groups and friendship circles [40,41]. Thus, to the extent that religious affiliation structures individuals' opportunities for interracial contact and friendship, it may affect their neighborhood racial-composition preferences.

### *2.2. Racial Attitudes*

In addition to shaping the composition of individuals' social networks, religious traditions provide individuals with cultural tools that they use to organize experiences and interpret reality [42]. For many Americans, beliefs and assumptions rooted in their religious faith are central to informing their views of the social world, including race. The key to understanding how cultural tools acquired through religion can impact racial attitudes is to recognize that tools or schema are transposable [43]. That is, they are transposed or extended beyond the context in which they were learned to new and diverse situations. Differences in racial attitudes between religious groups may translate into real differences in neighborhood racial-composition preferences. Numerous studies have linked whites' negative stereotypes about and negative attitudes toward minorities to a stronger preference for same-race neighbors [9-13].

Because religion is so central to the lives of many evangelical Protestants, Emerson and Smith [20] contend that three features of their cultural "tool kit" directly shape their attitudes toward race and racial inequality: "accountable freewill individualism," "relationalism," and "anti-structuralism." These cultural tools, according to the authors, are rooted in evangelical Protestant theology. Theological understandings portray individuals as responsible for their own behavior and fate; the importance of a "personal relationship with Christ" for salvation is translated into emphasis on the potential positive or negative impact of interpersonal relationships; and claims that macro-level structural dynamics shape human outcomes are deemed incompatible with accountable individualism. According to this account, evangelical Protestants' cultural 'tool kit' both prevents them from acknowledging social structural causes of racial inequality and leads them to blame perceived dysfunctional social relations among blacks for their own disadvantage [20,30]. Emerson and Smith do not contend that an emphasis on individualism—and an accompanying wariness toward structural explanations of inequality—is unique to evangelical Protestants, rather that their culture and theology lead them to hold these beliefs more strongly than other white Americans.

A long tradition of research in social psychology has suggested that conservative Protestants, particularly fundamentalists, are racially prejudiced [44-46]. Emerson and Smith focus not on racial prejudice, however, but rather on how supposedly race-neutral beliefs drawn from their cultural 'tool kit' lead to problematic and inaccurate views of racial inequality. Greeley and Hout [47] similarly reject the notion that evangelicals are racially prejudiced, echoing earlier assertions that opposition to race-related policies may be based on "principled conservatism" [48]. Drawing on over twenty years of GSS data, Putnam and Campbell suggest that white evangelicals' racial attitudes have become less distinctive over time [29]. Tranby and Hartmann [49] offer an alternative view. As others have argued [50], they insist that conservative views of racial inequality and racial policy are not easily disentangled from racial resentment and anti-black bias. Based on a reading of Emerson and Smith's interviews with evangelicals, the authors argue that evangelical Protestants routinely engage in group-based negative stereotyping to explain racial inequality. Furthermore, because the norms and values that form evangelicals' idea of "American-ness" are implicitly white, they come to see demands for increased recognition of and assistance for minority groups as a threat [49].

Emerson and Smith's research has sparked a new debate about whether white evangelical Protestants' racial attitudes differ significantly from those of other whites [49,51-54]. Evidence for the distinctiveness of evangelical Protestants' attitudes is mixed, however, and these studies have varied widely in their methodological approaches to the question. Whether their racial attitudes are distinct from other whites depends both on how one measures evangelical Protestantism and to whom one compares them [55,56]. Emerson and Smith's work, for example, focuses on the roughly 8 percent of whites who self-identify as "fundamentalist," "evangelical," or "Pentecostal" *and* express a belief in the Bible and in an afterlife. Taylor and Merino [55,56] report that, even after controlling on background characteristics, these self-identified conservative Protestants are more likely than other religious groups to cite motivation or will power as reasons for black-white inequality and less likely to cite structural causes like discrimination or access to quality education. However, only in their high levels of opposition to spending on blacks do these Protestants show distinctive racial policy opinions. In contrast, the roughly 30 percent of whites whose denominational preference is evangelical Protestant are less distinctive in their racial attitudes. When region, education, and other background characteristics are controlled, these white evangelicals are statistically indistinguishable from mainline Protestants and Catholics in their explanations for racial inequality and differ on only one racial policy issue.

Other religious traditions may foster beliefs and attitudes that are more sympathetic toward racial and ethnic minorities. Scholars have noted that the stratification beliefs of white Protestants and Catholics are generally more individualistic and less structuralist than those of Jews, adherents of other non-Christian faiths, and the religiously unaffiliated [57,58]. Hunt [58] writes of a status hierarchy among religions, with Protestants and Catholics the dominant groups, other faiths and non-affiliates being "minority" religious traditions. Members of dominant groups may be more likely to be exposed to the dominant ideology regarding race and racial inequality. In contrast, religious groups outside the Protestant/Catholic mainstream are minorities of a sort and may share a "religious underdog" perspective that positively inclines them toward other "out-groups" [58] Indeed, Taylor and Merino [55,56] find that the primary attitudinal divide among whites is between Christian groups and the more racially progressive non-Christians. In sum, if religious tradition helps to shape white Americans' racial attitudes, it may also contribute to their neighborhood racial preferences, as numerous studies have linked negative stereotypes about and negative attitudes toward minorities to a stronger preference for same-race neighbors [9-13].

A number of socio-demographic characteristics have been linked to neighborhood racial preferences, including age, educational attainment, income, marital status, and size of community [9,12,13,59]. To the extent that white evangelical Protestants differ from whites in other religious traditions on these socio-demographic characteristics, they may be distinctive in their neighborhood racial preferences. Notable differences in socioeconomic status and educational attainment continue to exist between religious groups in the United States [60,61]. Individuals that grow up in evangelical denominations continue to attain lower levels of education than other whites, resulting in somewhat lower levels of education among white evangelicals as a whole [61]. In addition, marriage, childbearing, and homeownership tend to occur earlier for evangelical Protestants than for other groups [62-64]. Significantly, scholars have suggested that homeowners or those with children in the home hold a greater stake in their neighborhoods and may have particular preferences about neighborhood composition and quality [13,59]. Finally, white evangelicals are relatively concentrated in the South, where whites generally have more conservative racial attitudes. Carefully analyzing the relationship between religious affiliation and socio-demographic characteristics is critical to understanding how religion might affect racial attitudes. As discussed earlier, white evangelical Protestants' beliefs about racial inequality and their views of racial policies are largely indistinguishable from those of other Christians after accounting for their socio-demographic characteristics [56]. This study will examine the extent to which white evangelical Protestants differ from other whites on these socio-demographic characteristics and how any observed differences relate to their neighborhood racial preferences.

Using a uniquely suited survey item from the 2000 GSS, this study will examine the relationship between religious affiliation and neighborhood racial-composition preferences among non-Hispanic whites in the United States. The analysis will proceed in two major steps. First, it will examine the bivariate relationship between individuals' religious tradition and their preferred neighborhood racial composition. Second, using OLS regression, it will determine whether religious group differences in preferences persist after controlling for socio-demographic factors. In addition, the multivariate analysis will examine whether differences between religious traditions are attributable to differences in two key areas shown to be important for neighborhood racial preferences: stereotyping about and social isolation from racial and ethnic minorities.

### **3. Data and Methods**

The General Social Survey (GSS) is administered biannually to stratified, multi-stage samples of non-institutionalized English-speaking Americans over the age of 17 by the National Opinion Research Center (NORC) at the University of Chicago. The sampling technique is designed to identify a nationally representative sample of households. The key variables in the current study come from the "Multi-Ethnic United States" topical module administered on the 2000 GSS. The current study employs a sub-sample of non-Hispanic whites. The decision to limit the sample to non-Hispanic whites was guided by two main considerations. First, beginning with Emerson and colleagues' work [20,30], much of the recent research on religion and race has focused on white evangelical Protestants, including Blanchard's study linking the size of a community's white evangelical institutional base to its level of black-white residential segregation [23,51,52]. Second, because whites' majority status and avoidance of racially mixed neighborhoods are thought to contribute to and uphold residential segregation, much of the research on neighborhood racial preferences focuses on whites [8,26]. Understanding how religious affiliation shapes non-whites' neighborhood racial preferences is certainly worthy of attention, but beyond the scope of the current study. Table 1 contains descriptive statistics for the dependent and independent variables used in the current study.


**Table 1.** Descriptive Statistics for Dependent and Independent Variables.

### *3.1. Dependent Variable*

Participants of the "Multi-Ethnic United States" topical module were shown a card depicting a single house surrounded by fourteen other houses. They were instructed as follows: "Now I'd like you to imagine a neighborhood that had an ethnic and racial mix you personally would feel most comfortable in. Here is a blank neighborhood card, which depicts some houses that surround your own. Using the letters A for Asian, B for Black, H for Hispanic or Latin American and W for White, please put a letter in each of these houses to represent your preferred neighborhood where you would most like to live. Please be sure to fill in all of the houses." In the data set, each household is coded individually, allowing for a calculation of the racial and ethnic composition of respondents' preferred neighborhood composition. The dependent variable in the current study is the percentage of households that respondents filled in as "white." Preliminary analysis in SPSS indicated that OLS regression is appropriate to use in this case. Despite some clustering at one end of the distribution (roughly 19% of non-Hispanic white respondents prefer an all-white neighborhood), the mean (0.55) and median (0.50) are similar and the skewness (0.40) and kurtosis (−1.1) values are near zero.

### *3.2. Independent Variables*

Religious tradition is determined using the RELTRAD scheme that classifies individuals on the basis of their stated denominational preference into one of seven major categories: "Evangelical Protestant," "Mainline Protestant," "Black Protestant," "Catholic," "Jewish," "Other," or "None" [65]. The "Other" category is residual and includes adherents of Eastern religious traditions as well as several non-traditional Western traditions [66]. The very small number of non-Hispanic whites in the Black Protestant category necessitates its omission.

This study employs a racial stereotyping measure used in several prior studies of neighborhood racial preferences [8-13]. It is scaled from −6 to +6 and is constructed from five survey items in which respondents were asked to rate each of the four major racial or ethnic groups (white, black, Asian, Hispanic) on a given characteristic (intelligence, laziness, violence-prone, committed to strong families, committed to fairness and equality for all). High (positive) scores indicate unfavorable ratings of out-groups relative to one's own group; low (negative) scores indicate favorable ratings of out-groups; 0 indicates no perceived difference. Cronbach's alpha for the scale is 0.62.

Social isolation from racial and ethnic minorities is measured by three separate survey items in which respondents were asked: "Do you personally know any" "Hispanics or Latin Americans," "Blacks," and "Asian Americans?" Each item is coded such that "0" indicates that the respondent reports not personally knowing anyone from the group, while a score of "1" indicates that the respondent reports knows a member of the group. The current study also employs a measure of racial homogeneity in respondents' communities. Studies have indicated that the racial composition of individuals' current communities affects their neighborhood racial preferences [59]. Furthermore, spatial and social isolation from ethnic outgroups is associated with more negative outgroup perceptions [38]. Respondents were asked to estimate the "percentage of the people who live in your local community" that are white.

Socio-demographic variables include sex, age (in years), total family income, and years of education. Dichotomous variables indicate whether the respondent is married, is a homeowner, has children in the home, or lives in the South. In addition, this study uses the variable XNORCSIZ, which is a measure of the size of a respondent's place of residence, ranging from 1 (open country) to 10 (city greater than 250,000).

### **4. Results**

Table 2 compares non-Hispanic white evangelical Protestants to whites in other religious categories on key socio-demographic variables. Several significant differences are worth noting. Evangelical Protestants are most different from the non-Christian groups—Jews, adherents of "other" faiths, and the unaffiliated. Compared with these groups evangelicals are older, less educated, live in less populated areas, and are more likely to live in the South. Evangelicals tend to be older than Catholics and far more likely to live in the South. Compared with mainline Protestants, they are less educated and more likely to live in the South. The religiously unaffiliated are less likely than evangelical Protestants to be married or own their home.


**Table 2.** Comparison of Evangelical Protestants to Other Groups on Key Socio-Demographic Variables.

Source: 2000 General Social Survey; N = 878; \* Difference from evangelical Protestants statistically significant (p < 0.05, 2-tailed); \*\* p < 0.01, 2-tailed

Table 3 compares evangelical Protestants to the other religious categories on the focal independent variables in the current study: social isolation from minorities and racial stereotyping. Again, differences between evangelical Protestants and the three non-Christian groups are most notable. Compared with these whites, evangelical Protestants are significantly less likely to report personally knowing Hispanics or Asian Americans, and more likely to hold negative stereotypes about minorities. Evangelical Protestants show little distinctiveness from Catholics and mainline Protestant, however. The only significant difference is evangelicals' greater likelihood of knowing blacks compared with Catholics.


**Table 3.** Comparison of Evangelical Protestants to Other Groups on Key Independent Variables.

Source: 2000 General Social Survey; N = 878; \* Difference from evangelical Protestants statistically significant (p < 0.05, 2-tailed); \*\* p < 0.01, 2-tailed.

Table 4 shows the neighborhood racial-composition preferences of non-Hispanic whites by religious tradition. Evangelical Protestants prefer the most racially homogeneous neighborhood, with an average of roughly 60 percent white [67]. Mainline Protestants prefer only slightly more diverse neighborhoods, with an average of 57.7 percent. Catholics (53.5%) and Jews (53.7%) both prefer neighborhoods in which slightly over half of all households are white. Only those of "other" faiths and the unaffiliated prefer, on average, a neighborhood in which whites do not make a majority. Those of "other" faiths have the weakest preference for same-race neighbors with an average of 42.8 percent. The religiously unaffiliated prefer a neighborhood in which about 49 percent of households are white [68]. All groups except for the "other" faith group prefer a neighborhood that is between 15 and 17 percent black. Differences in percent Asian and Hispanic are a bit more noticeable. Evangelical Protestants prefer the lowest percentage of both Asians and Hispanics, followed by mainline Protestants. Those of "other" faiths and the unaffiliated prefer the highest percentage of Asians and Hispanics.

Table 5 contains results from an OLS regression analysis of the percentage of households that respondents filled in as white. Model 1 compares non-Hispanic white evangelical Protestants to whites in each other religious tradition [69]. Results reveal that, before accounting for socio-demographic characteristics, evangelical Protestants' preference for same-race neighbors is significantly greater than that of every religious group but mainline Protestants. Model 2 introduces socio-demographic variables. Results indicate that group differences in socio-demographic characteristics largely explain the distinctiveness of white evangelical Protestants' neighborhood preferences [70]. Net of these controls, evangelical Protestants' preference for same-race neighbors is significantly greater than only those in the "other" faith category.


**Table 4.** Neighborhood Racial Preferences among Non-Hispanic Whites, by Religious Tradition.

Source: 2000 General Social Survey; N = 878

**Table 5.** Preference for Same-Race Neighbors Among Other White Religious Groups Compared to White Evangelical Protestants.



**Table 5.** *Cont.*

Source: 2000 General Social Survey;

Unstandardized coefficients reported, standard errors in parentheses.

† p < 0.10; \* p < 0.05; \*\* p < 0.01

a Reference group is evangelical Protestants

*Note*: Models 2 and 5 also control for family income, sex, marital status, home ownership, and the presence of children aged 0–6 or 7–12 in the household. None of these variables approach significance, so for simplicity they are not reported.

Models 3 and 4 introduce measures of social isolation from racial and ethnic minorities, as well as negative stereotyping, in the absence of socio-demographic variables. Model 3 demonstrates that the measures of social isolation from minorities used in the current study only partly explain group differences in preferences. While differences between groups are diminished somewhat, evangelical Protestants still prefer a significantly higher percentage of white neighbors than do Catholics, those of "other" faiths, and the unaffiliated.

Likewise, Model 4 indicates that white evangelical Protestants' stronger preference for same-race neighbors is not well explained by a greater propensity to hold negative stereotypes about minorities. Again, differences between evangelicals and other groups shrink somewhat, particularly for the "other" faith category and the unaffiliated, but remain statistically significant. Model 5 presents results from the full model, which includes socio-demographic variables. The pattern of group differences in Model 5 is highly similar to the one in Model 2. While stereotyping and isolation from minorities partially explain evangelical Protestants' distinctiveness, Model 2 demonstrates that socio-demographic characteristics alone render insignificant the differences between evangelicals and all other groups but the "other" faith category. Overall, these results suggest that while socio-demographic characteristics explain white evangelical Protestants' stronger preference for same-race neighbors, racial stereotyping and racial isolation do not.

As other studies have found, several socio-demographic characteristics are significant predictors of neighborhood racial-composition preferences. Birth cohort is a strong predictor for non-Hispanic whites. In Model 2, each additional year of age is associated with a .4 percentage point increase in same-race households. Compared with non-Southerners, whites living in the South prefer a neighborhood with a significantly higher percentage of whites. Respondents from more populous areas have a weaker preference for same-race neighbors, as do more educated whites. Each additional year of education is associated with a nearly 2 percent decrease in preferred same-race neighbors. Household composition has an effect on preferences. In Model 2, those with children between 13 and 17 prefer fewer white neighbors. In the final model, however, residing in the South and having children between 13 and 17 are the only significant socio-demographic variables.

Generally speaking, non-Hispanic whites that report personally knowing members of minority groups have a weaker preference for same-race neighbors. However, compared with personally knowing African Americans, knowing Hispanics or Asian Americans has a stronger effect on neighborhood preferences. Net of controls for stereotyping and socio-demographic characteristics, only knowing Hispanics or Asian Americans is predictive of a weaker preference for same-race neighbors. Respondent's estimates of racial homogeneity in their current communities matter as well. The estimated percentage of whites in the respondents' community is significantly positively associated with a preference for same-race neighbors. Finally, like numerous other studies [9-13], this analysis finds that holding negative stereotypes about minorities is a strong predictor of neighborhood racial preferences for non-Hispanic whites. A one-unit increase on the stereotyping scale is associated with a roughly ten percent increase in percent white.

### **5. Discussion and Conclusions**

Prior research on the causes of racial residential segregation has focused on the role of economic and financial institutions and on the preferences of individuals regarding the racial composition of their neighborhoods. The role that social institutions, such as religion, might play in influencing intergroup relations and either isolating or integrating racial and ethnic groups has gone largely unexplored. Emerson and Smith [20] have argued that the segregation of American religion along racial lines contributes to the racialization of American society. This study examines the relationship between religious affiliation and neighborhood racial-composition preferences among non-Hispanic whites. Evangelical and mainline Protestants have the strongest preference for same-race neighbors, while those of various "other" faiths and the unaffiliated have the weakest. This finding closely mirrors the general pattern observed for a range of whites' racial attitudes, in which Protestants have the most conservative attitudes, religious minorities (Jews, other faiths, and the unaffiliated) have the most progressive, and Catholics are somewhere in between [55,56]. Interestingly, it also mirrors the racial diversity within each of these traditions, as Protestant congregations are the most racially homogeneous, followed by Catholic and non-Christian congregations [27,28]. Furthermore, this study finds that while socio-demographic characteristics largely account for white evangelicals' stronger preference for same-race neighbors, negative stereotyping and social isolation from minorities do not.

Emerson and Smith [20] have suggested that religion contributes to the racialization of American society by creating racially distinct social networks, thereby limiting opportunities for interracial bridging ties. In the current study, however, the extent to which respondents personally know members of minority groups only partly explained religious group differences in preferences. Furthermore, evangelical Protestants' concentration in the South and in less populated areas likely accounts for their lower likelihood of knowing Hispanics or Asian Americans, especially compared with Jews, those of "other" faiths, and the unaffiliated. It is possible that more detailed measures of interracial contact and social network composition may better explain religious differences in neighborhood preferences. Examining the role of interracial contact at places of worship may be a worthwhile strategy. There is evidence that such contact may be especially effective at improving racial attitudes [21].

Emerson and Smith argue that religion endows individuals with cultural tools that they use to interpret the social world. When applied to race, these cultural tools influence individuals' racial attitudes. However, in the current study, a measure of racial stereotyping did not account for white evangelical Protestants' stronger preference for same-race neighbors. Instead, controls for socio-demographic characteristics, particularly region and education, explained much of the distinctiveness of evangelical Protestants' preferences. It is possible that other measures of racial attitudes would better explain the religious gap in preferences. Additional research is necessary to determine how religious affiliation may affect racial attitudes, including preferences about neighborhood racial and ethnic diversity.

While Blanchard's study [23] found that the presence of mainline Protestant congregations is associated with lower levels of black-white residential segregation, the current study finds that mainline Protestants themselves are indistinguishable from evangelical Protestants in their neighborhood racial-composition preferences. Notably, Blanchard makes an *institutional* argument about the relationship between local congregations and interracial relations. Evangelical Protestant congregations tend to be less involved in providing community service and outreach than mainline congregations, thus missing opportunities to foster bridging social capital [35-37]. In addition, mainline clergy tend to be more liberal than the laity, which may be a contributing factor to the level and type of community involvement displayed by mainline congregations [71]. Despite the relatively conservative racial attitudes of their members, mainline congregations may nonetheless have a positive effect on community racial integration by fostering civic engagement and bridging social capital [72].

Emerson and colleagues' provocative publications [20,27,30] have spurred an important line of research about race and religion in the United States. However, their conclusions have not been universally echoed in other social science research [49,51,52]. This study and recent work by Taylor and Merino [55,56] indicate the need for important qualification to claims about the influence of religion on racial attitudes. Individuals whose denominational preference is evangelical Protestant have significantly more conservative racial attitudes than other white Americans and prefer more racially homogeneous neighborhoods. However, after accounting for their socio-demographic characteristics, this group loses much of its distinctiveness. This pattern of findings makes it less clear how religion influences whites' racial attitudes.

Why is it difficult to identify religious influences on racial attitudes? Perhaps because, as Bartkowski and Matthews suggest, "the very same constellation of religious beliefs and practices that can be used to eradicate racial stratification also can be enlisted to reinforce it" ([73], p. 164). For example, an evangelical theology that has been said to blind its adherents to structural racism and reinforce segregated churches and social networks also drives efforts at 'racial healing' and 'Christ-centered' race-bridging [74,75]. Such ambivalence is on display in Brown's [76] study using Detroit Area Studies data from the 1970s and 1990s. He reports that, despite their stronger denial of racial housing discrimination, white evangelical Protestants actually expressed significantly greater openness than other white Christians to living in racially integrated neighborhoods. Brown situates these interesting findings within Detroit's own history of race relations and religious activism. His study also highlights the need to be attentive to the differing religious dynamics within local communities.

Additional research is needed to determine how religion works to either inhibit or foster bridging ties across racial and ethnic boundaries. Blanchard's [23] 'closed community thesis,' posited as an explanation for higher levels of black-white residential segregation in evangelical-rich communities, warrants further testing. Blanchard's thesis draws on two important areas of inquiry in the sociology of religion. First, several studies have suggested that religious traditions vary in the extent to which their congregations facilitate civic engagement and the development of bridging social capital in the broader community [15,37,72,77,78]. Second, there is growing interest in how religious involvement and beliefs structure individuals' social networks and their opportunities for intergroup contact [29,32,34,79,80]. Future research should examine how involvement in congregations and their religious subcultures shapes both opportunities for and preferences regarding social connections across racial and ethnic lines, particularly when religious culture contributes to racial identity [81]. As Edgell and Tranby suggest, "if religious subcultures are shaped in the context of highly salient racial boundaries, they may in fact be about race" ([51], p. 284). In addition, the cultural tools individuals acquire through participation in religious subcultures color their experiences with racial and ethnic diversity [51,75,82,83]. The task of researchers will be to better understand how individuals draw on those cultural tools to bridge racial divides in their communities.

### **References and Notes**




### **Religion and Marriage Timing: A Replication and Extension**

### **Joshua J. Rendon, Xiaohe Xu, Melinda Lundquist Denton and John P. Bartkowski**

**Abstract:** Previous studies have revealed denominational subculture variations in marriage timing in the U.S. with conservative Protestants marrying at a much younger age than Catholics and the unaffiliated. However, the effects of other religious factors, such as worship service attendance and religious salience, remain overlooked. Informed by a theoretical framework that integrates the denominational subculture variation thesis and the gendered religiosity thesis, this study replicates, updates, and extends previous research by examining the effects of religiosity on the timing of first marriage among 10,403 men and 12,279 women using pooled cross-sectional data from the National Survey of Family Growth, 2006–2010. Our survival regression models indicate that: (1) consistent with previous research, Protestants in general, and conservative Protestants in particular, marry earlier than the religiously unaffiliated; (2) irrespective of denominational affiliation, increased frequency of worship service attendance decreases age at first marriage for both men and women, whereas religious salience is associated with earlier marriage only for women; (3) among Catholics, as worship service attendance increases, the waiting time to first marriage decreases; and (4) among Protestants, however, worship service attendance decreases age at first marriage for men who are affiliated with mainline and non-denominational Protestant churches, while for women the decrease in age at first marriage associated with worship service attendance is found for those who report a conservative Protestant affiliation. The complex intersections of denominational affiliation, frequency of worship service attendance, religious salience, and gender are discussed. Results suggest that religion continues to exert influences on marriage timing among recent birth cohorts of young Americans.

Reprinted from *Religions*. Cite as: Rendon, J.J.; Xu, X.; Denton, M.L.; Bartkowski, J.P. Religion and Marriage Timing: A Replication and Extension. *Religions* **2014**, *5*, 834–851.

### **1. Introduction**

The past several decades have witnessed a remarkable growth in scholarly research on marriage timing [1,2]. A particular stream of this burgeoning body of research has focused on the role of religion and documented denominational subculture variations in marriage timing in the U.S., with conservative Protestants marrying at a much younger age than Catholics and the unaffiliated [3–5]. Though previous studies have revealed notable religious subculture variations in marriage timing in the U.S., the effects of other religious factors such as worship service attendance and religious salience are understudied. Informed by a theoretical framework that integrates the denominational subculture variation thesis and the gendered religiosity thesis, this study replicates and extends previous research by examining the effects of religiosity on the timing of first marriage among recent birth cohorts of American young adults using pooled cross-sectional data from the National Survey of Family Growth, 2006–2010. In particular, this study addresses the following research questions: (1) Does religion continue to exert influences on marriage timing among recent birth cohorts of young Americans? (2) If so, are there denominational subculture variations as exhibited in previous research net of other religious factors? (3) Are worship service attendance (*i.e.*, religious network integration) and religious salience (*i.e.*, the internalization of religious norms and values) associated with marriage timing? (4) If so, do these associations vary across diverse faith traditions as expected by the denominational subculture variation thesis? (5) And finally, do these religious effects on marriage timing, if uncovered, vary by gender?

This study makes significant contributions to family and religious studies in several important ways. First, denominational subculture variations in marriage timing as reported by previous studies are largely based on survey data collected in the late 1980s and early 1990s. The present study explores whether these denominational subculture variations are still present among young Americans who were surveyed in the 21st century. This replication effort is warranted in light of recent social trends. Young Americans are increasingly experiencing multiple life course transitions in union formation, especially pre-marital cohabitation, which is known to delay entry into first marriage [6]. In addition, religiosity in terms of worship service attendance, prayer, belief in afterlife, and scriptural literalism (or scriptural inerrancy) among American adults has steadily declined across birth cohorts [7]. Given the trends toward delayed marriage and away from religious involvement, it is important to examine whether or not previously documented relationships between religion and marriage timing are still at work in the population.

Second, there is a lack of critical and rigorous validity checks in previous studies on denominational subculture variations in marriage timing. Previous studies examining the influence of denominational affiliation on marriage timing used a measure of denominational affiliation during adolescence [3]. While the use of denominational affiliation in adolescence meets the criterion of causal reasoning, adolescent religious affiliation may not truly reflect individuals' religiosity at the time of first marriage. In effect, it may reflect individuals' parental expectations and/or family religious traditions [8,9]. Given these oversights, this study extends the denominational subculture variation thesis by using denominational affiliation in both adolescence and adulthood. This comparative and fuller approach provides an ideal test for the robustness of the denominational subculture variation thesis in the context of change and continuity in religiosity across the individuals' life course.

Third, the effects of worship service attendance and religious salience on marriage timing have been understudied. In fact, one of the frequently cited studies even overlooked other religious factors including worship service attendance and religious salience [5]. The current study fills this research void by investigating: (1) denominational subculture variations in marriage timing net of worship service attendance and religious salience; (2) independent or net effects of worship service attendance and religious salience on marriage timing after controlling for denominational affiliation; and (3) intersectional or multiplicative effects of worship service attendance and/or religious salience and denominational affiliation on marriage timing.

Finally, the current study explores the effects of religiosity on marriage timing by gender. Though previous studies have examined the links between religion and marriage timing separately by gender [3,5], no explicit and systematic efforts have been made to understand gender differences in marriage patterns. Informed by previous scholarship that deems both religion and family as gendered institutions [10–12], this study examines the gendered effects of religious denominational affiliation, worship service attendance, and religious salience on marriage timing.

### **2. Review of Literature and Research Hypotheses**

The denominational subculture thesis was initially developed by religion scholars to: (1) rank-order religious denominations along a liberal-moderate-conservative continuum; and (2) assess the consistency of these rankings across a range of "pro-family" issues, with special attention to both between-denominational differences and within-denominational homo/heterogeneity [13]. Utilizing this theoretical perspective, recent scholarship has highlighted distinctive denominational subculture variations in marriage timing. Mormons (the Latter-day Saints), moderate Protestants, and conservative Protestants marry earlier than Jews or their unaffiliated counterparts. Catholics fall right in the middle of this marriage-timing spectrum by differentiating themselves from the early marrying Latter-day Saints and conservative Protestants and the late marrying unaffiliated and Jews [3,5].

These denominational differences in marriage timing are often accounted for by their distinctive subculture variations in theological beliefs and religious norms pertaining to pro-family attitudes, fertility patterns, gender differences in educational attainment and labor force participation, as well as gender ideologies [3,5]. Mormons and conservative Protestants, for example, place a primacy on marriage and family life and emphasize family roles as a source of sanctification and fulfillment. The subcultural emphasis on traditional family life encourages and supports marriage at younger ages [3,5]. The average age of first marriage for Catholics falls between Protestants and Mormons on one end and Jews and the unaffiliated on the other. Like Protestants, Catholics also espouse a pro-family theology that might lead to earlier marriage. However, the fact that the average age of first marriage among Catholics is later than that of Protestants may be related to the contours of the Catholic respondents. Catholicism is viewed by many as an integral part of their cultural and family identity. These individuals may identify themselves as Catholic on a survey even if they are not religiously engaged, thereby being called "cultural Catholics." Individuals who were raised in a Protestant tradition but are no longer religious are less likely to maintain their Protestant identity and therefore more likely to move into the unaffiliated category. As a result, the Catholic category of respondents includes larger numbers of nominally religious respondents as compared to Protestants, and therefore Catholics may be less distinct from the religiously unaffiliated than Protestants [8].

Though the denominational subculture variation thesis is informative in identifying and explicating the multifaceted linkages between religion and marriage timing, this line of research is not without limitations. After carefully reviewing this body of literature, several weaknesses are noteworthy. First, previous studies relied heavily on respondents' denominational affiliation during adolescence [3,5]. This operationalization practice makes sense in temporal order but can be problematic and misleading. It has been argued that adolescent denominational affiliation may not accurately reflect individuals' religious identities and commitment as their religious identities and beliefs continue to be shaped and reshaped by their own discoveries as they age [8,9]. Therefore, denominational affiliation during one's upbringing or adolescence can result in inconsistent and inaccurate measures of the subcultural contexts that influence the marriage timing of young adults. To rectify this research limitation, current denominational affiliation or, more ideally, denominational affiliation at first marriage should be used to serve as a critical check.

Second, by default, the denominational subculture thesis is predicated on the theological beliefs and religious norms of the religious traditions. As such, it overlooks possible denominational variations in other measures that gauge either public or private religiosity. Two such measures that can potentially affect marriage timing across various denominational families are worship service attendance and religious salience. As a measure of public religiosity, frequency of worship service attendance can shorten marriage timing in three significant ways: (1) those who attend worship services frequently can regularly receive a moral proclamation of the importance of marriage and other pro-family, pronuptial, and pronatal messages and teachings; (2) frequent attendance at worship services can provide opportunities to cultivate religious capital or networks through which one can interact with co-religionists to enhance their views of marriage and/or to dissuade or sanction those who stray from the religious teachings; and (3) frequent attendance at worship services can also serve as an indication of religious commitment, particularly commitment to marriage and family life. In a similar fashion, religious salience, as a measure of private religiosity, can affect marriage timing as well. Religious salience is a subjective measure of how important religion is to a person and the extent to which they have internalized the religious norms, values, and teachings of their religious community [14]. Individuals who report high levels of religious salience are more likely to internalize and adopt their religion's norms and values pertaining specifically to marriage and family life. They may also be more inclined to consult or use religious teachings to inform major life decisions. Religious salience is less tangible than worship service attendance in terms of religious commitment and the sacrifice of time, energy, or income. However, religious salience represents similar commitment based on subjective assessment of internalized importance of religion. There are good reasons to believe that both frequent worship service attendance and heightened religious salience can affect marriage timing directly. The mechanism of influence for each factor, however, is dependent upon the religious context where the individual is interacting with co-religionists or internalizing religious teachings. As such, attendance and salience may affect marriage timing differentially across denominational families because of different levels of strictness in religious ideologies, expectations, and practices. For example, these religious effects can be stronger for conservative Protestant denominations for their higher levels of biblical literalism, theological conservatism, and more frequent religious service attendance [13].

A third limitation of previous research is that in spite of the widely recognized gender differences in marriage timing and religiosity [5,10–12,15], prior studies on marriage timing have taken these differences for granted. Little, if any, attention has been given to the gendered effects of religion on marriage timing. This oversight is unfortunate because both marriage and religion have long been viewed as gendered institutions [10–12]. With reference to gendered marriage, the most widely canvassed explanation offered by Jessie Bernard is that there are two different marital realities, his and hers, such that marriage benefits husbands more than wives [16]. Recent scholarship continues to document gendered boundaries, segregated roles, and gender-differentiated meanings in marriage [10]. Turning to the gendered character of religion or religiosity, scholars of religion have concluded that women are more religious than men on a wide range of measures [11,12]. Scholars of religion offer three types of explanations for this gendered religiosity: (1) women are psychologically or naturally more inclined towards religion (psychological explanation); (2) women are predisposed to such religious values as nurturance, submission, and gentleness during their childhood socialization (socialization explanation); and (3) women's structural locations in society, such as childrearing roles, lack of labor force participation experiences, and their prioritization of family life, lead to a stronger religious orientation than their male counterparts (social location explanation). These two bodies of literature jointly suggest that denominational subculture differences and other religious variations in marriage timing can be different across the two gender groups.

Guided by the literatures reviewed above, the following hypotheses are formulated:

Hypothesis 1: Those who are affiliated with a faith tradition are more likely to marry at a younger age than those who are unaffiliated (Hypothesis 1A). Moreover, among Protestants, conservative Protestants will display the earliest entrance into marriage, followed by mainline Protestants (Hypothesis 1B). Catholics, on the other hand, are expected to be between the early marrying conservative Protestants and the late marrying religiously unaffiliated (Hypothesis 1C). Finally, denominational affiliation in adolescence will exhibit weaker and inconsistent subculture influences on marriage timing than denominational affiliation at young adulthood (Hypothesis 1D).

Hypothesis 2: Regardless of denominational affiliation, there will be a negative relationship between frequency of service attendance and marriage timing such that more frequent worship service attendance will be significantly associated with a younger age at first marriage (shorter waiting time to first marriage).

Hypothesis 3: Regardless of denominational affiliation, those who deem religion important in their lives will marry earlier than those for whom religion is viewed as unimportant.

Hypothesis 4: Worship service attendance and religious salience will affect marriage timing differently across faith traditions, with the strongest effects being observed for conservative Protestants.

Hypothesis 5: Given women's higher levels of religiosity and a stronger orientation towards family life, the religious effects on marriage timing will be stronger for women than for men.

### **3. Research Methods**

### *3.1. Data*

To test the hypotheses delineated above, this study used data from the 2006–2010 cycles of the National Survey of Family Growth (NSFG 2006–2010). These surveys were designed to provide reliable national data on cohabitation, marriage, divorce, remarriage, contraception, infertility, and the health of women and infants in the United States. The pooled NSFG 2006–2010 sample was nationally representative of the civilian, non-institutionalized population, consisting of 10,403 men and 12,279 women ages 15–44. The NSFG has consistently surveyed young Americans aged 44 or younger because of its focus on reproductive health. This age truncation may limit the estimation of the religious effects on marriage timing due to its disproportionate inclusion of unmarried young respondents, thus hampering the potential to generalize the study findings to other populations (e.g., older populations). In spite of this limitation, however, the NSFG contains excellent life course transition questions pertaining to cohabitation and marriage, making the data suitable for the current study. In addition, given the analytical focus of this study on early marriage among young Americans, especially those who are religious, the pooled NSFG data are well suited for this purpose.

The NSFG 2006–2010 used a complex survey design and oversampled underrepresented groups, including African Americans and Hispanics. The survey was conducted by the Institute for Social Research (ISR) at the University of Michigan, from June 2006 through June 2010 under contract from the National Center of Health Statistics. The merged public-use data and the codebook were downloaded from the Centers for Disease Control and Prevention by the first author.

### *3.2. Dependent Variable: Waiting Time to First Marriage*

Taking a cue from previous research, this study used an event history approach to analyze the survey data [5]. Within this analytical framework, the dependent variable was conceptualized and operationalized as the waiting time to first marriage, which was constructed via two different procedures. First, for respondents who were ever married, the waiting time to first marriage was calculated by subtracting date of birth from date of first marriage (in century month = year × 12 + month). Second, for respondents who were unmarried, their waiting time was calculated by subtracting date of birth from date of interview (in century month calculated similarly as before). Respondents who were unmarried at the time of interview represent censoring cases in this study, which is one of the major advantages of using event history methods for data analysis. In other words, those who were not married at the time of interview will not be excluded from the current study because they may marry at a later time. As displayed in Table 1, the average waiting time for women is 24.4 years and for men 25.2 years, respectively, a year older for women and 3 years younger for men compared to Xu *et al.*'s study [5]. It is worth noting that one of the striking differences in this sample from Xu *et al.*'s study is the larger number of individuals who reported never being married at the time of interview. In this sample, approximately 55% of women and 64% of men reported never being married as opposed to 21% of women and 27% of men in their study based on the first wave of the National Survey of Families and Households.


**Table 1.** Sample Characteristics by Gender.


**Table 1.** *Cont.* 


**Table 1.** *Cont.* 

### *3.3. Key Covariates: Religious Variables*

In the current study, religious denominational affiliation, frequency of worship service attendance, and religious salience were used as covariates to replicate, update, and extend previous research on marriage timing. Consistent with previous research, denominational affiliation reported by respondents was employed to operationalize denominational subculture variations [5,13]. But due to possible changes in religiosity over the life course of respondents [8,9], two versions of the religious denominational affiliation variables were used: (1) respondents' denominational affiliation as an adolescent and (2) respondents' current (at the time of study) denominational affiliation. Because the National Center of Health Statistics did not release the original denominational affiliation variables with detailed denominational membership, the denominational affiliation variables available in the public use data filewere pre-collapsed, thus incongruent with previous studies that utilized detailed denominational membership. These variables were dummy-coded into five broader faith traditions: Catholic, conservative Protestant (Baptists and other fundamentalist Protestants), mainline Protestant (Methodists, Lutherans, Presbyterians, and Episcopal groups), other Protestant (non-denominational or Protestant groups not listed in the survey), and other religions (Muslims, Jews, Latter-day Saints and Jehovah Witnesses) with the unaffiliated serving as the reference group.

The NSFG 2006–2010 also included frequency of worship service attendance and religious salience. Worship service attendance was recorded as an ordinal variable to indicate respondents' public religiosity, with seven response categories ranging from 1 = "never attend" to 7 = "attend more than once a week." For ease of interpretation in the models, worship service attendance was treated as a continuous measure (a categorical version of the variable was experimented but no difference surfaced). While previous research used frequency of worship service attendance at age 14 [3], this study made use of current worship service attendance instead because of excessive missing data in the adolescent worship service attendance variable (missing data were observed for the vast majority of respondents).

Finally, the NSFG 2006–2010 included religious salience. This measure was used to gauge respondents' private religiosity. The NSFG 2006–2010 asked how important religion was in respondents' daily life, which was dummy-coded with 1 = salient ("very important" and "somewhat important") and 0 = not salient ("not important"). The category of "not salient" was used as the reference.

### *3.4. Other Covariates: Control Variables*

To conduct the statistical analysis, the following control variables (covariates) were included to avoid possible spurious effects of religiosity on marriage timing. Race/ethnicity was dummy-coded into Black, Hispanic, and other race/ethnicity, with white serving as the reference category. Premarital cohabitation was also dummy-coded into 1 = "ever had premarital cohabitation" and 0 = "never had premarital cohabitation" (the reference category). Current educational attainment, in actual years, was dummy-coded into two variables: high school and more than high school with less than high school serving as the reference group. Because respondents' employment status at time of marriage was unavailable, current employment status was used and dummy-coded with 1 = employed and 0 = otherwise. Family structure at age 14 was used and dummy-coded into 1 = biological two-parent family and 0 = other family arrangement. Current family resources were measured by whether the family received public assistance, which was dummy-coded into 1 = "yes" and 0 = "no". Since region of residence was not provided in the public use data, metro statistical area was used and dummy-coded into 1 = urban and 0 = rural to control for marriage market differences. Finally, years of survey were dummy-coded into three variables: 2007, 2008, and 2010, with 2006 serving as the reference category.

### *3.5. Analytic Strategies*

Following previous studies, the effects of the religious variables on marriage timing were analyzed by using a series of log-logistic parametric survival models (selected as the best fitting model among five different types of parametric survival models; not shown but available upon request). This modeling strategy has several advantages, including but not limited to: (1) censored observations for those who were not married at the time of study were incorporated into the analysis; (2) the waiting time to first marriage with flexible distributions was accounted for; and (3) fuller information was used for statistical modeling; and (4) a direct comparison with previous studies, such as Xu *et al.*'s study published in 2005, was possible.

To test study hypotheses, a nested modeling technique was used such that Model 1 was a replication model that included denominational affiliation and all of the control variables. Each model was run once with the adolescent affiliation variables and once with the current affiliation variables in order to allow a comparison of the two different measures of denominational affiliation. Models 2 and 3 were extension models that included worship service attendance and religious salience, respectively, while controlling for denominational affiliation and other covariates. Model 4 was the full model that combined all religious variables. It is important to note that these models were estimated separately for men and women in order to explore gender differences. In addition, the effects of worship service attendance and religious salience, along with statistical controls, were estimated separately for each of the five denominational families by gender. In essence, these models assessed complex moderating or intersectional effects of denominational subculture, worship service attendance or religious salience, and gender on marriage timing. It should be noted that all of these models were estimated by using the complex survey and multiple imputation procedures in Stata 13 to adjust for design effects and missing values [17,18].

### **4. Results**

### *4.1. Denominational Subculture Variations in Marriage Timing*

Model 1 of Tables 2 and 3 show general support for Hypothesis 1A. Consistent with previous studies, the negative and significant regression coefficients displayed in survival regressions indicate that the waiting time until first marriage was shorter for those who were affiliated with any faith tradition than for those who were unaffiliated. In other words, religiously affiliated respondents were more likely to marry at a younger age. This pattern generally holds for both denominational affiliation during adolescence and adulthood as shown in Tables 2 and 3.

While all religious groups marry younger than the unaffiliated, Hypothesis 1B suggests that conservative and mainline Protestants will report the earliest entries into marriage, respectively. As shown in the tables (Models 2–4), net of worship service attendance, religious salience, and other statistical controls, conservative Protestants exhibit the most consistent and early marrying effects (the negative and significant regression coefficients are observed across both versions of the denominational affiliation variables). These findings offer partial support for Hypothesis 1B pertaining to distinctive conservative Protestantism. However, in contrast to Hypothesis 1B, mainline Protestants do not marry significantly earlier than the unaffiliated once all of the covariates are controlled for in the models. Hypothesis 1C statesthat Catholics will fall between the early marrying conservative Protestants and late marrying unaffiliated in marriage timing. As it turns out, Hypothesis 1C is supported only for current denominational affiliation but rejected for adolescent denominational affiliation (no statistical differences between Catholics and the unaffiliated are observed) if other religious factors are not considered (Model 1). Once additional religious factors are added to the models, however, there is no longer a significant difference in the marriage timing of Catholics and the religiously unaffiliated.

Hypotheses 1A–1D examined the denominational subculture variations in marriage timing. The results show that both adolescent and current religious affiliations are related to marriage timing, but as was expected, current religious affiliation is a stronger measure of the subcultural influences on marriage timing (somewhat consistent with Hypothesis 1D). In addition, the subcultural context that appears to have the most consistent and significant influence on early entry into marriage is that of conservative Protestantism.

Although we found some support for denominational subculture variation, some of these variations in marriage timing are mediated by both worship service attendance and religious salience (the ancillary regression analyses indicating significant mediating effects are not shown here but available upon request). In the case of current denominational affiliation for women, denominational subculture variations in marriage timing are completely mediated (or explained) away by religious salience (see Models 3–4 of Table 2). That is, after controlling for either worship service attendance or religious salience, denominational subculture variations in marriage timing become considerably weaker or even completely nullified as compared to one of the earlier studies [5] (once again, the ancillary regression analyses confirmed these significant mediating effects).


**Table 2.** Maximum Likelihood Parameter Estimates from Log-Logistic Survival (AFT) Regressions of Waiting Time on Religious Variables for Women.

Note: \* *p* < .05; \*\* *p* < .01; \*\*\* *p* < .001. Race/ethnicity, premarital cohabitation, education, employment, family structure at age 14, poverty, urban-rural residence, and year of study are statistically controlled.

### *4.2. Worship Service Attendance, Religious Salience and Marriage Timing*

Hypothesis 2 predicts that worship service attendance will be related to a shorter waiting time to first marriage. Models 2 and 4 in Tables 2 and 3 provide the opportunity to test this hypothesis as an extension of previous studies. As expected, the negative and statistically significant survival regression coefficients show that after controlling for denominational affiliation and other covariates, every unit increase in frequency of worship service attendance is associated with a younger age at first marriage (a shorter waiting time to first marriage). This robust pattern holds for both men and women, which strongly supports Hypothesis 2. Turning to religious salience as addressed in Hypothesis 3, Models 3 and 4 in Tables 2 and 3 suggest that those for whom religion was deemed salient married earlier than those who viewed religion as unimportant in their lives. The effects of religious salience are far more robust for women than for men. In Model 4, for example, religious salience is no longer statistically significant for men when worship service attendance is included in the model, but it continues to be significant for women. In light of these results, Hypothesis 3 is partially supported.


**Table 3.** Maximum Likelihood Parameter Estimates from Log-Logistic Survival (AFT) Regressions of Waiting Time on Religious Variables for Men.

Note: \* *p* < .05; \*\* *p* < .01; \*\*\* *p* < .001. Race/ethnicity, premarital cohabitation, education, employment, family structure at age 14, poverty, urban-rural residence, and year of study are statistically controlled.

Turning to Hypothesis 4, we examine whether worship service attendance and religious salience operate differently across the different religious traditions in the study. Table 4 shows the coefficients for each of these two religious variables when the models are run separately by denomination and gender. The frequency of worship service attendance is systematically and negatively associated with time to first marriage for both male and female Catholics. Among Protestant groups, on the other hand, the negative association between worship service attendance and length of time to first marriage is not consistent across all of the models. Among conservative Protestants, there is a significant coefficient for attendance for females who were conservative Protestants in adolescence and those who are currently conservative Protestants. For the men, however, the relationship between attendance and marriage timing is only significant for those men who were conservative Protestant during adolescence. Among mainline Protestants, it is only the men for whom more frequent worship service attendance is significantly related to earlier marriage. Religious salience is significantly related to a shorter time to first marriage only for female mainline Protestants and male Catholics. Thus, the moderating effects involving religious salience are minimal and not systematic. Taken together, the survival models featured in Table 4 lend some credence to Hypothesis 4 pertaining to worship service attendance. It appears that the way in which worship service attendance influences marriage timing does vary across religious traditions. And while the effect of attendance for conservative Protestants was expected, we also found that attendance has a strong influence on marriage timing among Catholics. The support for Hypothesis 4 related to religious salience is generally weak and in most cases, statistically trivial.


**Table 4.** Maximum Likelihood Parameter Estimates from Log-Logistic Survival (AFT) Regressions of Waiting Time on Religious Variables By Denomination and Gender.

Note: \* *p* < .05; \*\* *p* < .01; \*\*\* *p* < .001. Race/ethnicity, premarital cohabitation, education, employment, family structure at age 14, poverty, urban-rural residence, and year of study are statistically controlled.

### *4.3. Gender Differences*

As was predicted in Hypothesis 5, several gendered religious effects on marriage timing emerged from this study. The noteworthy results can be stated as follows: (1) denominational subculture variations in marriage timing (especially conservative Protestant affiliation) are more prominent and systematic for men than for women after controlling for worship service attendance and religious salience; (2) religious salience, a measure of private religiosity, is more robust in affecting marriage timing for women than for men; and (3) for each of the five denominational families as depicted in Table 4, the early marrying effects of worship service attendance are more pronounced for men than for women. Based on these findings, there are noticeable gender differences in religious effects on marriage timing as was expected in Hypothesis 5. It appears, however, that private religious salience is more of an influence for women, while marriage timing among men is more consistently influenced by public religious practice (worship service attendance).

### **5. Discussion and Conclusions**

This study was designed to replicate, update, and extend previous research on the relationship between religion and marriage timing using the National Survey of Family Growth, 2006–2010, a nationally representative sample of young American men and women. As anticipated, a series of multivariate survival regression models revealed important religious effects on the waiting time to first marriage. In the pages that follow, these findings are summarized and highlighted.

First, consistent with Hypothesis 1A, respondents who were affiliated with all faith traditions exhibited shorter waiting time to first marriage than those who were unaffiliated. This finding undergirds the notion that pro-family and pro-marriage values that characterize virtually all faith traditions continue to play an important role in marriage timing. As pointed out by Uecker and Stokes [19], religion is responsible in part for much of the early marriages in recent birth cohorts. However, it is worth noting that denominational subculture variations documented in this study are less robust than previously reported, especially after other religious factors are simultaneously considered.

In support of Hypothesis 1B, conservative Protestants showed the most robust and consistent early marrying effects compared to the unaffiliated, which was followed by other Protestants (men). Given conservative Protestants' enthusiasm for pro-family values, traditional gender ideologies, and family life, these findings are highly anticipated. They underscore the subcultural uniqueness associated with conservative Protestantism that is highlighted by their distinctive biblical literalism and theological conservatism. On the other hand, inconsistent with Hypothesis 1C, those who were affiliated with the Catholic faith tradition were not statistically different from the late marrying unaffiliated once other religiosity measures (worship service attendance and religious salience) were included in the analysis. This result is consistent with the earlier discussion regarding the cultural identity of many Catholics who may identify as Catholic even thought their religious engagement more closely resembles the religiously unaffiliated. This conclusion is further supported by the findings regarding worship service attendance in Table 4. In the model for Catholics we found that those who did attend regularly were more likely to reflect the pro-family and pro-marriage stance of the Catholic Church with a shorter waiting time to marriage. Furthermore, it was difficult to conclude firmly if Hypothesis 1D was supported or rejected because the two versions of denominational affiliation exhibited different patterns in their effects on marriage timing across the two gender groups. But it seems safe to conclude that the results derived from current denominational affiliation tell "more interesting" stories.

Second, this study concluded that irrespective of denominational affiliation, as worship service attendance increased, the waiting time to first marriage became shortened, which lent strong credence to Hypothesis 2. In fact, attendance at religious services emerged as the most robust predictor of marriage timing. Moreover, worship service attendance also acted as a mediator, explaining away not only some of the denominational affiliation effects but also the effects of religious salience, which was particularly pronounced for men. These mediating effects are theoretically important for two reasons: (1) religious denominational affiliation can be nominal such that its effects on marriage timing will not matter unless it is manifested through religious practice, such as worship service attendance; and (2) in line with previous research on the linkages between religion and family life, public religiosity often exerts more pronounced net effects on marital dynamics, relationship quality, and other dimensions of marital well-being [20].

Partially consistent with Hypothesis 3, religious salience was also found to shorten the waiting time to first marriage. However, there were striking gender differences. For men, religious salience lost its statistical significance when worship service attendance was introduced, whereas for women religious salience was statistically important throughout the analysis. So why does the internalization of the religious pro-family and pro-marriage orientations and teachings matter for women but not for men? On the one hand, religious institutions tend to be gender-stratified, such that women have fewer opportunities to exhibit their religiosity publicly other than attendance at religious services. As a result, women tend to internalize their faith through such private acts as prayers and scripture studies. On the other hand, men have abundant opportunities to externalize their religious faith by serving as leaders or teachers, thus private religiosity seems to matter less for men than for women.

In partial support of Hypothesis 4, this study indicated that worship service attendance and religious salience affected the waiting time to first marriage differently across denominational families with the effects of religious salience being far less systematic than worship service attendance. Attendance at worship services mattered more consistently for Catholics than for various Protestant groups even though Catholics were not that different from the unaffiliated in marriage timing as reported above. So while Catholics as a group are not significantly different from the religiously unaffiliated, Catholics who attend regularly do in fact marry at a younger age than their less-attending or non-attending counterparts. This within-group heterogeneity was also noted for conservative Protestant women and mainline or other Protestant men. These patterns of within-group religious heterogeneity complement nicely the denominational subculture variations observed in marriage timing.

In general, this study found some evidence to support Hypothesis 5. The gendered effects of religious salience as a predictor and mediator for women, and the denominational specific effects of worship service attendance for men, supported the contention that like the institution of family or marriage, religious institutions are also gendered. In the context of marriage timing, these results echo broader forms of gender segregation in society by the well-known differentiation between the public (worship service attendance for men) and private (religious salience for women) spheres of life for both sexes.

While this study yielded some interesting and important findings, several research limitations and directions for future research need to be addressed and discussed. As noted previously, future research should utilize more refined religious denominational groups, which are less likely to be available in the public use data. Failure to separate denominational groups, such as the Latter-day Saints and Jews, from other faith traditions can make the interpretations difficult. As such, access to the original data is essential to use an appropriate classification scheme to group denominational families. In addition, with a growing number of Americans being self-classified as unaffiliated [21], further distinction of the unaffiliated group becomes necessary in order to examine properly the effects of this group on marriage timing. It is important to note that the unaffiliated group can consist of atheists, agnostics, and others who may classify themselves as unaffiliated because they tend to come from an inter-faith home, thus embracing different beliefs, values or norms. Furthermore, due to a large amount of missing data, the variable of worship service attendance in adolescence could not be used in this study. Attention is needed in future research to better record respondents' retrospective responses. An event history calendar can be very helpful in probing and recording respondents' past religious practices. Likewise, in this study many covariates serving as statistical controls were not measured at first marriage. Instead, they were measured at the time of interview. As a result, no causal relationships and implications are suggested.

Additionally, as noted previously, the National Survey of Family Growth focuses on a young population with a narrow age range from 15 to 44. Given the increasing age at first marriage across the population [22], the current data include significant numbers of respondents who are not yet married. While these factors limit our ability to measure the eventual marriage patterns of these respondents, the data do allow us to examine the marriage patterns of young adults and the prevalence of early marriage within the population. Finally, we suggest that future research incorporate qualitative studies, which can help better understand the nuanced motivations or desires for earlier or later entrance into marriage.

In closing, this study makes several noteworthy contributions to family and religious studies. In spite of the declining religiosity across birth cohorts in the U.S. and the increasing age at first marriage in recent decades [7,22], this study documented continued and important impacts of religion on marriage timing among young Americans. Echoing previous research, the present study observed persistent denominational subculture variations in marriage timing, especially for conservative Protestants compared to other denominational groups. In addition to these notable religious subculture variations, this study also revealed accelerating effects of religious attendance and salience on marriage timing. While the impacts of religious attendance were noted for both men and women, the effects of religious salience were particularly pronounced for women. This finding pertaining to private religiosity underscores the gendered nature of both family and religious life in contemporary America. This gendered finding is also nicely complemented by the complex intersection of gender, denominational affiliation, worship service attendance, and religious salience, suggesting that future research should move beyond the denominational subculture variation thesis and bring gender into the study on religion and family life in general and religion and marriage timing in particular. Finally, it is recommended that similar theoretical and methodological approaches used here be considered to examine additional life course transitions such as the timing of premarital and post-divorce cohabitation, divorce, and/or remarriage.

### **Author Contributions**

JJR and XX conceived of the study and preformed the statistical analyses. MLD and JPB provided theoretical guidance and helped to draft the manuscript. The authors jointly edited and approved the final manuscript.

### **Conflicts of Interest**

The authors declare no conflict of interest.

### **References**

