Religiosity, Spirituality, and Attitudes Toward Same-Sex Marriage

Attitudes toward same-sex marriage have changed dramatically over the last decade. U.S. adults are becoming more supportive of same-sex marriage, and there are a number of reasons for this change. Our research examines the relationship between cohort, religiosity, spirituality, and attitudes toward same-sex marriage. Using data from the 2012 and 2014 General Social Surveys, we examine the differential impact of religiosity and spirituality by cohort on attitudes toward same-sex marriage. We present models for four separate cohorts: The Millennials, Generation X, the Baby Boomers, and the Silent Generation. The Millennial cohort exhibits significant differences from the other birth cohorts. The results of our analyses locate various changes in these attitudes and provide directions for future research.


Introduction
Attitudes toward same-sex marriage have changed dramatically over the last decade. U.S. adults are becoming more supportive of same-sex marriage, and there are a number of reasons for this increase. In the 1988 General Social Survey (GSS), only 12.5% of the American public agreed or strongly agreed that same-sex marriages are acceptable (Smith, Marsden, & Hout, 2015); however, by 2003, one third of the American public supported same-sex marriage (Pew Research Center, 2013). Today, public opinion polls show that half or more of the public support same-sex marriage (Pew Research Center, 2013;PollingReport.com, 2014;Smith et al., 2015). These polls also show that support for same-sex marriage has increased across all birth cohorts/age groups over the last decade (e.g., Smith et al., 2015). Another indicator of the change in cultural outlooks is the number of states that now recognize same-sex marriage as legal. Currently, 14 states have not legalized same-sex marriage. However, that number is about to change as the U.S. Supreme Court ruled that same-sex couples have a right to marry anywhere in the United States.
Although the ruling by the U.S. Supreme Court may suggest that research which examines support for same-sex marriages is now moot, there are strong reasons to advocate additional research on this volatile and timely social issue. With the exception of legal abortion, few social issues have been more divisive and garnished more media attention than same-sex marriage. In addition, there are several issues such as workplace rights and family law which will likely emerge among gay-rights advocates and become a lightning rod for social debate and political decision making. For example, private-and religious-based schools may lose their taxexempt status should they refuse to alter their policies against same-sex marriage. Moreover, the judicial decision reached by the Supreme Court will surface in public debates among political candidates at the state and federal levels. Hence, explicating the variables that influence attitudes regarding same-sex marriage remains an important topic for social science research.
Our research focuses on differences in attitudes toward same-sex marriage as they are shaped by religiosity and spirituality across birth cohorts. The impact of religion on issues related to sexuality and family needs little documentation (e.g., Sherkat & Ellison, 1999), while the interest in spirituality is a more recent topic among scholars (Hill et al., 2000). Religiosity refers to an individual's level of emersion within a religious community-often defined by the liturgical doctrines of the denomination. Spirituality connotes an outlook that is grounded, not in religion, but in a distinct approach to the world that eschews established organizational ties (Chaves, 2011). We elect to examine attitudinal variations by birth cohorts because they have been shown to be important predictors for attitudinal outlooks regarding sexuality and family (Lewis & Gossett, 2008;Smith, 1994). Prior to 1988, few studies examined the religious, spiritual, and secular correlates of public attitudes toward same-sex marriage (Sherkat, Powell-Williams, Maddox, & de Vries, 2011). The GSS included the question about same-sex marriage in 1988 and has in every survey since 2004. In the last decade, other national probability surveys such as the Pew Research Center have included similar questions. We use the 2012 and 2014 GSS years because they include items that tap religiosity, spirituality, and attitudes toward same-sex marriage and afford the opportunity to examine differences across younger and older birth cohorts. Although it would be highly useful to examine the role of religiosity and spiritualism as they affect the mundane experiences of individuals who must interpret and incorporate these values into their daily lives (Walls, Woodford, & Levy, 2014), our research relies on survey data that cannot tap into this aspect pertaining to same-sex marriages. However, our analysis does point to an important component of the cultural resources that establishes the basis for an individual's syncretism (Woodford, Walls, & Levy, 2012).
The trends in attitudinal change are not evenly distributed across birth cohort/generation, religious identification, religious participation, belief, or subjective spirituality. Recent research (Jones, Cox, & Navarro-Rivera, 2014;Taylor, 2014) shows significant differences in outlooks by cohort over the last decade. Current research identifies four cohorts for comparisons of social attitudes and behaviors: (a) the Millennial cohort, (b) Generation X, (c) the Baby Boom cohort, and (d) the Silent Generation (e.g., Bengtson, Putney, & Harris, 2013;Jones et al., 2014;Taylor, 2014). These researchers identify the Millennial cohort as people who were born between 1981 and 1996, Generation X as those who were born between 1965 and 1980, Baby Boomers as those who were born between 1946 and 1964, and the Silent Generation as people who were born before 1945. We follow this operationalization for our cohort comparisons of attitudes toward samesex marriage.
The purpose of our study is to examine cohort comparisons of religiosity and subjective spirituality on attitudes toward same-sex marriage in three ways. First, we examine mean differences in attitudes among the four cohorts in the analysis with and without controls. Second, we regress attitudes toward same-sex marriage on birth cohort, religiosity, spirituality, and control variables for the entire sample. Third, we run separate regression analyses to assess the relationship between religion and spirituality on attitudes toward samesex marriage within each cohort controlling for sociodemographic variables.

Religiosity, Spirituality, and Same-Sex Marriage
Religious intensity and commitment can be manifested in a number of ways. Religiosity involves a number of elements or facets of one's life, and there are multiple aspects as to what the term or concept conveys (see Ellison, Gay, & Glass, 1989;Glock & Stark, 1965). The literature examining the relationship between religiosity and social attitudes and behaviors has a long history in social science research. Recent research demonstrates that various aspects of religiosity impact attitudes toward same-sex marriage (Jones et al., 2014;Olson, Cadge, & Harrison, 2006;Sherkat et al., 2011). Personal identity with a religious tradition, denomination, or family of denominations reinforces personal outlooks of what is viewed as acceptable or unacceptable behavior. Van Geest (2007) suggests that conservative evangelical Protestants are more likely to see same-sex marriage as a moral issue that runs counter to their value systems. Mainline Protestants are much more in favor of gay rights (Van Geest, 2008). Some research indicates attitudes of Catholics appear to be becoming more in line with mainline Protestants (Sherkat, de Vries, & Creek, 2010), even though others (e.g., Olson et al., 2006) find differences. Hence, identity with a religious tradition likely affects attitudes toward same-sex marriage and may vary by birth cohort. Steensland et al. (2000) is currently the most widely used religious categorical scheme and is the operationalization for affiliation in the present study. Public religious participation through attendance at religious services reinforces these views. Attendance at religious services promotes a social network of individuals who are likely to share particular outlooks. Belief in the literal interpretation of sacred writings and subjective practices such as prayer are also important components of religiosity. Recent research reports a significant relationship between the subjective components of religiosity and attitudes toward same-sex relations (e.g., Barringer, Gay, & Lynxwiler, 2013;Ford, Brignall, VanValey, & Macaluso, 2009;Olson et al., 2006).
In addition, subjective spirituality can be an alternative to religious influences on attitudes and behaviors. Even though social scientists have addressed subjective spirituality in the cohort and religion literature (e.g., Roof & McKinney, 1987;Wuthnow, 2007), little research has examined the relationship between subjective spiritual views and attitudes toward sexuality and same-sex marriage. This study operationalizes spirituality as spiritual but not religious (SBNR). Research consistently recognizes a relationship between spirituality and religiosity (e.g., Ammerman, 2013;Hill et al., 2000). However, a significant percentage of the population conceptualizes spirituality as separate from religiosity and that is the operationalization for the current analysis. Recent research is beginning to examine the impact of SBNR identity on attitudes toward same-sex sexuality (e.g., Barringer et al., 2013). Our research extends this line of inquiry by addressing the possible influence of SBNR on attitudes toward same-sex marriage and whether the influence varies by birth cohort.

Religious Affiliation
Recent research documents substantial religious denominational differences in attitudes toward same-sex sexuality (Cadge, Olson, & Wildeman, 2008;Finlay & Walther, 2003;Gay, Ellison, & Powers, 1996;Gay & Lynxwiler, 2010;Loftus, 2001) and same-sex marriage (Burdette, Ellison, & Hill, 2005;Jones et al., 2014;Sherkat et al., 2011). Personal identity with a religious group is likely important in attitude formation. For some time, research has assumed that religious denominations are the "locus of subcultural differences" (Jeffries & Tygart, 1974, p. 318). Even though the term subculture carries varied connotations, it usually refers to norms and practices that are shared by members of the group that are distinct from the larger population. Hence, attitudes toward same-sex sexuality and same-sex marriage are likely to vary by religious affiliation and its corresponding ideology. Recent research addresses the role denomination or religious tradition plays in influencing attitudes and behaviors concerning the morality of same-sex sexuality (Cadge et al. 2008;Kenneavy, 2012;Olson & Cadge, 2002;Sherkat et al., 2010;Van Geest, 2007). Such studies show that evangelical and sectarian religious traditions are the least tolerant of same-sex relations. A major question in the literature continues to be how we categorize religious affiliation (Gay & Ellison, 1993;Roof & McKinney, 1987;Steensland et al., 2000). Currently, the most widely used religious categorical scheme is the one developed by Steensland et al. (2000), and it is the scheme used in the present study. The Steensland et al.'s typology of religious traditions is operationalized as Catholic, Jewish, mainline Protestant, evangelical Protestant, African American Protestant, "Other" affiliates, and unaffiliated/no religious preference. An important aspect of this study is the variation of the religious tradition effect by cohort. Unfortunately, due to sample size issues for certain religious groups in our society (e.g., members of the Jewish faith, Mormon respondents, Muslims), our analysis is restricted to the largest religious traditions in the GSS data. They are Catholics, mainline Protestants, evangelical Protestants, African American Protestants, and no religious preference respondents. Hence, the study focuses on Protestants, Catholics, and no preference respondents.

Public Religious Participation
Individuals who participate in public religious activities are more likely to incorporate the ideas expressed in the larger group. Attendance at religious services is the structured activity that affords the most available route to public participation. Attendance exposes the individual to norms and practices of the group and reinforces attitudes and behaviors associated with a particular religious collectivity (Ellison et al., 1989;Jeffries & Tygart, 1974;Sherkat et al., 2011). Evangelical religious traditions are the most likely to be intolerant of same-sex relations (Gay et al., 1996;Hertel & Hughes, 1987;Walls et al., 2014;Woodford et al., 2012) while the more mainline Protestant denominations have been more supportive of same-sex relations (Kenneavy, 2012;Sherkat et al., 2010). Even though attendance at religious services or public participation may reinforce the variations in attitudes by religious tradition, the bulk of the literature reports that attendance overall reduces the support for same-sex marriage (Becker, 2012;Cochran, Chamlin, Beeghley, & Fenwick, 2004). Extending this line of inquiry, we examine the relationship between attendance at religious services and attitudes toward same-sex marriage for each of the four cohorts.

Biblical Literalism and Prayer
Two subjective aspects of religiosity are beliefs and prayer. One of the most important aspects of the relationship between religious belief and attitudes toward same-sex marriage in the United States is a belief in the literal interpretation of the Bible. For this study, the Bible is the primary sacred text for the traditions in American mainline religion. Biblical literalism is the belief that the words in the Bible were inspired by God and are to be interpreted literally. This interpretation of the Bible is most commonly found among evangelical denominations of Christianity, although it is not exclusive to them (Burn & Busso, 2005). Even though research reports substantial heterogeneity within religious denominations, evangelical and sectarian Protestant groups are much more likely to interpret same-sex sexuality as morally wrong (Kenneavy, 2012;Knust, 2011).
Prayer is also a significant component of subjective religiosity as individual religious behavior. Roof and McKinney (1987) refer to individualistic phenomena as the "new voluntarism" that leads to enhanced privatization of faith that is more anchored in personal realms. Using the 2014 GSS, a simple cross tabulation of frequency of attendance and frequency of prayer shows that roughly 30% of respondents who never attend religious services pray at least once a day. As Wuthnow (2007) points out, "prayer is fairly common among the American public" (p. 128). Americans pray for all kinds of reasons both internal and external, contrition and thanksgiving, and personal and collective. Sociological research uses frequency of prayer as an independent variable (Ellison, Bradshaw, Flannelly, & Galek, 2014) and a dependent variable (Baker, 2008). And some religious professionals encourage the use of prayer to search for a "life partner" under certain circumstances. Finally, recent research (Barringer et al., 2013) shows that the frequency of prayer is related to attitudes toward same-sex relations. Hence, prayer is included in our analysis of attitudes toward same-sex marriage.

Subjective Spirituality
Subjective spirituality may have a different and more subtle effect on attitudes toward same-sex marriage. It is a more difficult concept to measure than religiosity and is influenced by a variety of personal and societal factors (Ammerman, 2013). Spirituality is not necessarily institutionally grounded and is subject to individual interpretations. It is more personal in definition (Roof, 1993(Roof, , 1999. While studies consistently show that the majority of Americans report that they are both spiritual and religious (Marler & Hardaway, 2002), roughly 15% of the 2012-2014 GSS respondents can be classified as SBNR. Heelas and Woodhead (2005) suggest that Americans are increasingly using spirituality as a guide for understanding in their everyday lives.
It is beyond the scope of this article to provide a definition of "spirituality." The GSS does not ask respondents what spirituality means to them. Some social scientists suggest that religion and spirituality are distinct concepts (Schlehofer, Omoto, & Adelman, 2008). Even though research clearly shows that the two are not mutually exclusive (Marler & Hardaway, 2002), a number of surveys (e.g., GSS) demonstrate that a significant percentage of respondents report that they are spiritual in one question but not religious in another. While we may not know their subjective definitions, we do know that the combination of the questions indicate that they feel they are spiritual, but they do not feel they are religious. Recent research reports that these SBNR attitudes are related to favorable outlooks toward same-sex sexuality (Barringer et al., 2013). This may very well extend to attitudes toward same-sex marriage. Those who are SBNR are less likely to attend religious services or identify with a particular denomination, and hence less likely to be influenced by a particular religious organization's stance on same-sex relations. Recent research shows that respondents identifying as SBNR varies across birth cohorts .

Cohort, Religiosity, Spirituality, and Attitudes Toward Same-Sex Marriage
As noted earlier, we are interested in the relationship between cohort, religiosity, spirituality, and attitudes toward same-sex marriage. Ryder (1965) is most frequently cited as the researcher to emphasize cohort in the study of changes in attitudes and behavior. Since this seminal work 50 years ago, cohort comparisons have been used to examine a wide range of attitudes and behaviors. Many of these studies use an Age/ Period/Cohort design. However, this study is limited to cohort comparisons at one point in time because the data are cross-sectional and some questions of interest are only asked in certain years. Nevertheless, we can identify different birth cohorts even though they are at different ages in 2012 and 2014. As noted earlier, support for same-sex marriage has increased for all birth cohorts (Pew Research Center, 2015;Smith et al., 2015). Whether or not attitudes change over the life course is not our research question. Moreover, the issue of same-sex marriage was not part of public discussion several years ago. Why emphasize cohort when only cross-sectional data are available? Birth cohorts were socialized in different time periods and cultural epochs depending on the time they were born. Baby Boomers experienced different religious, cultural, and institutional aspects of society than the Millennials. Generation X lived through different experiences than the Silent Generation and so on. As a result, the determinants of attitudes may vary across birth cohort regardless of their current age. Therefore, we are interested in more than simple age effects. Other researchers have also examined cohort effects using GSS data (e.g., Sherkat et al., 2010;Sherkat et al., 2011). We want to extend the literature by examining the potential differential effects of religiosity and spirituality on attitudes toward same-sex marriage by cohort.
We follow the operationalization of birth cohorts used in current social scientific studies (Jones et al., 2014;Pew Research Center, 2013;Taylor, 2014). These surveys show that younger birth cohorts (i.e., the Millennials and Generation X) are more accepting of same-sex marriage than their older cohort counterparts. But it is more than just a simple younger versus older distinction. Each of these polls shows that support for same-sex marriage across cohorts is a consistent, albeit unsurprising, pattern. The Millennials show the highest level of support followed in order by Generation X, the Baby Boomers, and the Silent Generation. While this could be an age effect, it may also reflect a cohort effect. This finding in and of itself does not tell us anything new. However, little research has compared the determinants of attitudes toward same-sex marriage within birth cohort. And while research has examined the relationship between religiosity and cohort (Hoge, Johnson, & Luidens, 1994;Wuthnow, 2007;Smith, 2009), very little research addresses the relationship between religiosity, spirituality, cohort, and attitudes toward same-sex marriage. Should Millennials and Generation X carry their current attitudes toward religion, same-sex marriage, and other social attitudes and behaviors throughout their life course, our society will experience significant change on several dimensions.
Millennials are more likely to report no religious preference, and they attend religious services at lower levels than all other cohorts. They pray less frequently and have the lowest percentage of respondents who hold a literal interpretation of the Bible . Millennials have the lowest proportion of respondents who identify with denominations that are categorized as evangelical or mainline Protestants. The percentage who identify as Catholics is very similar to all other birth cohorts. For the most part, all indicators are that they are less religious. Do these trends in religiosity and spirituality across cohort affect attitudes toward same-sex marriage?
The purpose of this research is to take a step in that direction. Toward that end, first we examine mean differences in attitudes among the four cohorts in attitudes toward samesex marriage with and without controls. Second, we regress attitudes toward same-sex marriage on birth cohort, religiosity, spirituality, and control variables for the entire sample. Third, we analyze separate regression analyses to assess the relationship between religion and spirituality on attitudes toward same-sex marriage within each cohort controlling for sociodemographic variables.

Control Variables
Cohort differences in attitudes toward same-sex marriage may be related to factors other than religious and spiritual matters. Recent research demonstrates the importance of controlling for a variety of sociodemographic variables (e.g., Becker, 2012;Cadge et al., 2008;Lannutti & Lachlan, 2007;Sherkat et al., 2011). This research demonstrates that attitudes toward same-sex marriage vary by marital status, gender, socioeconomic status, geographic residence, and race and ethnicity to mention a few. As a result, this study incorporates control variables in the analysis. Previous research has demonstrated that women are more supportive of samesex sexuality and same-sex marriage. Higher educational attainment has also been associated with support for samesex marriage. Sherkat et al. (2010) reports variations in attitudes by race, region of residence, and community size. People who live in the south and reside in rural areas are generally less supportive of same-sex marriage. Baunch (2012) points to the importance of including African Americans in an analysis of attitudes toward same-sex marriage and Ellison, Acevedo, and Ramos-Wada (2011) report the importance of the variation among Hispanics. It is possible that the inclusion of African American Protestants will affect the results for other African Americans. It is also likely that evangelical Protestants will affect the Hispanic effect. Ellison et al. (2011) point to the strong opposition by evangelical Latinos as contributing to considerable heterogeneity of attitudes toward same-sex marriage. Given the presence of race and ethnicity in the literature, dummy variables to represent African American and Hispanic respondents are incorporated in the analysis. In short, drawing from previous research, we also included a dummy variable for female respondents, dummy variables for divorced/separated and never married respondents, a dummy variable for southern residence, a Likert-type scale item for community size, and a 20-point scale for years of educational attainment.

Analytic Strategy
The analytic strategy used ordinary least squares (OLS) multiple regression to examine the effects of birth cohort, religiosity, spirituality, and sociodemographic variables on attitudes toward same-sex marriage. A dummy variable for survey year was included to assess whether survey year affected the analysis. The inclusion of the survey year dummy variable did not affect the analysis in any significant way. That is, the inclusion of the variable did not affect the significance level of any of the independent or control variables.

Method
Like many previous studies of differences in religious, social and political attitudes, data for this study were taken from the GSS. The data in these surveys were collected from nationwide samples through interviews (Smith et al., 2015). For the following analysis, only respondents who were interviewed in 2012 or 2014 were included. The GSS is an appropriate data set because the data set contains survey items tapping attitudes toward same-sex marriage, religious and spirituality items, and a wide range of sociodemographic and background characteristics of respondents. A total of 2,977 respondents were asked the question concerning same-sex marriage. Our analysis focused on religious traditions identified by Steensland et al. (2000), and as a result, not all respondents in the total sample are included in the analysis. As we analyzed birth cohorts separately, some groups were omitted due to small sample sizes. For example, the total sample only included 40 respondents who identified as Jewish and 18 who identified as Muslim. Moreover, there were other religious affiliations that were not part of the operationalization of religious traditions incorporated in the analysis due to limited sample size. Furthermore, one of the difficulties with the GSS data was identifying Asian Americans and other races and ethnic groups. Of the total respondents on the race question, 458 were coded "other" (White and Black are the other two categories). The Hispanic question for ethnic origin was more direct. Finally, with the listwise deletion of data for the analysis, the final sample size is 2,375. See the appendix for the total frequencies for all variables and categories included in the analysis.

Dependent Variable
The item addressing attitudes toward same-sex marriage was present in both the 2012 and 2014 GSS. Respondents were read the following statement: "'Homosexual' couples should have the right to marry one another." The respondents were then presented five response categories that ranged from 1 (strongly disagree) to 5 (strongly agree). All other responses (i.e., "can't choose" and "no answer") are excluded from the analysis.

Independent Variables
Birth cohort. Four birth cohorts were identified for the analysis. The Millennial cohort was identified as respondents who were between 1981 and 1996 (n = 544). Generation X was identified as respondents who were born between 1965 and 1980 (n = 653). Baby boomers were identified as respondents who were born between 1946 and 1964 (n = 785), and the Silent Generation were those who were born before 1945 (n = 393). The baby boom cohort served as the reference category.
Religious affiliation. Religious affiliation was measured using two questions in the GSS. The first question was, "What is your religious preference? Is it Protestant, Catholic, Jewish, some other religion, or no religion?" An additional question asked whether the response to the first question was Protestant. The question was, "What specific denomination is that, if any?" These two questions were used to recode religious affiliation in accordance with the Steensland et al. (2000) categorical scheme. Hence, not all religious affiliations were included in the analyses. In addition, to run separate models for each birth cohort, Jewish respondents were omitted due to small sample sizes within cohorts. The resulting religious categories were mainline Protestants, evangelical Protestants, African American Protestants, Catholics, and no religious preference. Mainline Protestants served as the reference category.
Attendance at religious services. Public religious participation was measured by religious attendance. The question was, "How often do you attend religious services?" The responses ranged from 0 (never) to 8 (more than once a week). Only valid responses were used in the analysis. That is, "don't know" and "no answer" responses were omitted from the analysis.
Biblical literalism. Literalism was measured through the following question to assess their beliefs concerning the Bible. "Which of the following statements comes the closest to describing your feelings about the Bible-'The Bible is the actual word of God and is to be taken literally'; 'The Bible is the inspired word of God but not everything in it should be taken literally, word for word'; or 'The Bible is an ancient book of fables, legends, history, and moral precepts recorded by man?'" Respondents who considered the Bible to be "the actual word of God and is to be taken literally" were coded (1) and all others were coded (0) for the analysis.
Prayer. The final variable that addresses subjective religiosity was the question concerning prayer. The frequency of prayer is measured by the following question: "About how often do you pray?" Responses were coded (1) never, (2) less than once a week, (3) once a week, (4) several times a week, (5) once a day, and (6) several times a day. "Don't know" and "no answer" responses were omitted from the analysis.
SBNR. We use two items from the GSS to operationalize the SBNR variable and followed the measurement strategy of previous research (e.g., Chaves, 2011). The first question addressed subjective spirituality. The question in the GSS is, "To what extent do you consider yourself a spiritual person? Are you . . . " The possible responses ranged from 1 (very spiritual) to 4 (not spiritual at all). Subjective religiosity was measured by a question tapping the importance of religion in their everyday lives. The question in the GSS was, "To what extent do you consider yourself a religious person? Are you . . . " The possible responses to this question ranged from 1 (very religious) to 4 (not religious at all). "Not applicable," "don't know," and "no answer" responses for both of these questions were omitted from the analysis. An SBNR dummy variable was created through a combination of the subjective spirituality and religiousness questions. Individuals who report being "very spiritual" or "moderately spiritual" on the spirituality variable and "slightly religious" or "not religious at all" on the religious variable were coded (1) and represent the SBNR respondents. All other valid responses were coded (0). "Don't know" and "no answer" responses were omitted from the analysis.

Control Variables
Marital status and gender. The marital status question in the GSS asks respondents if they are currently married, widowed, divorced, separated, or have ever been married. Marital status was recoded to represent three statuses. Dummy variables were created to represent respondents who were married or widowed, divorced or separated, and never married. Married respondents served as the reference category in the analysis. Gender was coded (1) to represent female respondents and males were coded (0).

Southern residence and community size.
A dummy variable was created for southern residence. The GSS includes an item that indicates the respondent's area of residence. The coding followed the U.S. census coding for region. The resulting codes in the GSS are (1) New England, (2) Middle Atlantic, (3) East North Central, (4) West North Central, (5) South Atlantic, (6) East South Central, (7) West South Central, (8) Mountain, and (9) Pacific. A dummy variable was created using the South Atlantic, East South Central, West South Central codes to represent the South (South = 1, all others = 0). Urban residence was measured using the Survey Research Center (SRC) Belt Code (Survey Research Center, University of Michigan). The variable was recoded so that the "central city of the 12 largest SMSAs" was coded (6), "central city of the remainder of the 100 largest SMSAs" was coded (5), "suburbs of the 12 largest SMSAs" was coded (4), "suburbs of the remaining 100 largest SMSAs" was coded (3), "other urban" was coded (2), and "rural" was coded (1).
Educational attainment. The responses for educational attainment in the GSS were in years of school completed and ranged from 0 to 20. Using years of education as proxy for educational attainment was relatively straightforward for years 0 through 12. Years of education after high school may be more subjective. That is, for many people, it may take longer than 2 years to obtain an associate degree or longer than 4 years to get a bachelor's degree and so on. The assumption is that the number of years beyond high school reflects the appropriate vocational and academic degrees.
Race and ethnicity. Dummy variables were created for race and ethnicity by using the questions concerning race and Hispanic identification. For purposes of this analysis, respondents who identified as White, African American, or Hispanic (regardless of race) were included. White respondents served as the reference category. Table 1 displays the means, standard deviations, and proportions for attitudes toward same-sex marriage, the religious variables, subjective spirituality, and the control variables for each cohort in the analysis. The table also presents the sample size for each of the cohorts. Hence, proportions in the table reflect the number of respondents represented by each dummy variable. The table exhibits some interesting comparisons on the dependent and independent variables across the four cohorts. First, an ANOVA revealed significant cohort mean differences for attitudes toward same-sex marriage, F(3, 2371) = 55.01, p < .001. Post hoc comparisons for mean differences indicated that all were statistically different from one another (all ps < .001). There was also a pattern of mean differences. A steady decrease in level of acceptance was evident across the four cohorts. The Millennial cohort had the highest mean score (more likely to agree that same-sex couples should be allowed to marry) and the Silent Generation had the lowest.

Results
Second, the table shows both religious affiliation category differences and similarities. The proportion of each cohort who was Catholic was relatively constant. The same was the case for African American Protestants. The cohort differences were among no preference respondents, evangelical Protestants, and mainline Protestants. Table 1 shows that the proportion of no religious preference respondents was highest among the Millennial cohort and lowest among the Silent Generation. This pattern was reversed for both evangelical and mainline Protestants. The Millennials had a lower proportion of these two religious categories than the older cohorts.
Third, the Millennial cohort had the lowest mean and proportion scores for the remaining three religiosity variables (i.e., attendance, Biblical literalism, and frequency of prayer). The Silent Generation had the highest mean for attendance at religious services (M = 4.08, SD = 2.95) highest proportion of Biblical literalist (.35), and the highest mean for the frequency of prayer (M = 4.55, SD = 1.55). Even though research shows that attendance has a tendency to increase with age regardless of birth cohort (Newport, 2012), others (Miller & Nakamura, 1996) argue that older cohorts have attended religious services more frequently across the life course. In addition, the Millennial cohort attends religious services less than the baby boom cohort did at the same age. Bengtson et al. (2013) contend that the oldest birth cohort in this study is the most religious among the cohorts identified, and the results in Table 1 are consistent with their research. We also examined mean differences across religious affiliation categories for the total sample. An ANOVA revealed significant religious affiliation mean differences for attitudes toward same-sex marriage, F(4, 2371) = 81.23, p < .001. With one exception, all post hoc mean comparisons were statistically significant (p < .001). The lone exception was the difference between mainline Protestants and African American Protestants. Hence, these two religious categories appear to be more similar than dissimilar in their attitudes toward same-sex marriage. This is important to note because mainline Protestants represent the reference group for all regression models in Tables 2 and 3.
The subsequent analyses addressed the relative impact of religiosity, spirituality, and sociodemographic variables on attitudes toward same-sex marriage for the full sample (Table  2) and for each cohort (Table 3). Table 2 displays two models where attitudes toward same-sex marriage were regressed on three dummy variables for cohort (Model I) and on the cohort dummy variables, religious and spirituality independent variables, and controls (Model II). The first model included dummy variables for three of the birth cohorts and accounted for 6.5% of the variation in attitudes toward same-sex attitudes, F(3, 2371) = 54.69, p < .001. Model 1 shows the regression coefficients for the Millennial cohort, Generation X, and the Silent Generation. The Baby Boom cohort was the reference group for both models. The coefficients represent mean differences (Model I) and adjusted mean differences (Model II) between the three dummy variables representing the Millennials, Generation X, and the Silent Generation and the Baby Boom reference cohort. Model I shows that the Millennials (b = .710, p < .01) and Generation X (b = .319, p < .01) had higher scores than the Baby Boomers on the attitudes toward same-sex marriage scale. In contrast, the Silent Generation had a lower score (b = −.451, p < .01) than the Baby Boomers. The coefficients in Model II control for religious and spirituality independent variables and the sociodemographic control variables. The model explained 32.4% of the variation in attitudes toward same-sex marriage, F(20, 2354) = 56.33, p < .001. The results for the birth cohort dummy variables mirrored the results without controls. That is, cohort differences remained when religious, spirituality, and control variables were included in the full model. Table 2 also displays the results for the religious and spirituality variables for the total sample. Model II shows that evangelical Protestants were significantly less supportive of same-sex marriage than mainline Protestants (b = −.260, p < .01). Catholic respondents (b = .222, p < .01) and no preference respondents (b = .197, p < .05) were more supportive than the comparison category. African American Protestants were not significantly different from mainline Protestants. Table 2 shows that as attendance at religious services increased, support for same-sex marriage decreased. This is also the case for frequency of prayer. Respondents who believe in the literal interpretation of the Bible were less supportive of same-sex marriage than those who do not hold those beliefs. The final independent variable in Model II represents the SBNR respondents. Those who were SBNR had a higher score on the attitudes toward same-sex marriage scale than all others. This is consistent with previous research on the relationship between subjective spirituality and attitudes toward samesex relations (Barringer et al., 2013).
The sociodemographic control variables showed a pattern that is generally consistent with the extant literature. Females were more supportive than males, never married respondents were more supportive that married respondents, and southerners were less supportive than respondents in other parts of the country. In addition, as community size increased, so did support for same-sex marriage, and higher educational attainment predicted greater support. The variable representing African American respondents was not statistically significant and neither was the variable representing Hispanic respondents. Thus, the analysis of the full sample provided a baseline for cohort differences, the impact of religiosity and spirituality, and other determinants of attitudes toward samesex marriage. All but one of the cohort differences remained when controls were included. We extend the existing literature on cohort and religiosity differences by analyzing each cohort independently. Table 3 displays the results of the regression analysis for each cohort. That is, separate regression analyses were run for Millennials, Generation X, Baby Boomers, and the Silent Generation. These regression analyses afforded the opportunity to examine the unique determinants of attitudes toward same-sex marriage by cohort. Each model includes the religious, spirituality, and control variables. The first model shows the coefficients for the Millennial cohort and accounted for 25.4% of the variation, F(17, 526) = 10.52, p < .001. The analysis for Millennials shows that Catholics (b = .391, p < .01), African American Protestants (b = .552, p < .01), and no preference respondents (b = .429, p < .01) were more supportive of same-sex marriage than mainline Protestants. While other research points to some variation within religious tradition, this research also points to differences between religious traditions among young adults even when control variables are included in the models (Walls et al., 2014). Interestingly, there was no significant difference between evangelical Protestants and mainline Protestants. Like the model for the full sample in Table 2, respondents who believe in the literal interpretation of the Bible were less supportive. In addition, as the frequency of prayer increased, support for same-sex marriage decreased. In contrast to other cohorts, attendance at religious services had no effect on attitudes toward same-sex marriage for Millennial respondents. SBNR respondents also showed no significant difference.
In the Generation X model of Table 3, we see a different pattern of results. This model accounts for 30.6% of the variation in attitudes, F(17, 635) = 16.49, p < .001. Evangelical Protestants were significantly less supportive than their mainline Protestant counterparts (b = −.422, p < .01) while African American Protestants and no religious preference respondents showed no significant differences. Catholic respondents were marginally significant (b = .266, p < .10). This pattern was very close to the reverse of the pattern found for the Millennial cohort. Like the Millennial cohort, as attendance at religious services increased, support for same-sex marriage decreased. The effects of the belief in the literal interpretation of the Bible and the frequency of prayer were consistent with the Millennial cohort and the analysis of the full sample. However, unlike the Millennial cohort, respondents identified as SBNR were more supportive of same-sex marriage than the reference category (b = .277, p < .01).
The third model in Table 3 presents the results for the Baby Boom cohort and accounts for 30.9% of the variation in attitudes toward same-sex marriage, F(17, 767) = 20.15, p < .001. Unlike the other cohorts, there were no religious affiliation differences. Like Generation X, as attendance at religious services increased, support for same-sex marriage decreased, and those who believe in the literal interpretation of the Bible were less supportive of same-sex marriage. Frequency of prayer was not statistically significant. However, like Generation X and unlike the Millennial cohort, Baby Boomers who were identified as SBNR showed more support for same-sex marriage (b = .393, p < .01).
The last model in Table 3 displays the results for the Silent Generation and accounts for 31.8% of the variation in attitudes toward same-sex marriage, F(17, 375) = 10.29, p < .001. This cohort was the least supportive of all cohorts in general. The model for the Silent Generation shows that evangelical Protestants were significantly less supportive than the reference category (b = −.507, p < .01) and no religious preference respondents were significantly more supportive of same-sex marriage (b = .625, p < .01). Catholics and African American Protestants showed no significant differences from the mainline Protestant reference category. There were only two religious or spiritual variables in the model that were significant. Attendance at religious services had a marginally significant negative effect (b = −.048, p < .10). And respondents who believe in a literal interpretation of the Bible were less supportive of same-sex marriages (b = −.189, p < .01). There was no significant effect for respondents who were identified as SBNR.
Among the control variables, female respondents were consistently more supportive than males across all cohorts. This is also the case for educational attainment. For all cohorts, as educational attainment increased, so do attitudes toward same-sex marriage. There was no consistent pattern for the other control variables. The effect of southern residence was found among Baby Boomers but not among other cohorts. This finding raises some interesting questions concerning southern residence.

Limitations
First, our study uses a cross-sectional design, and we are somewhat limited as a result. We are unable to examine cohorts when they were at the same age (e.g., Millennials and Baby Boomers when they were both in their twenties), we could not follow a cohort over time, and the data are simply not available to assess period or time of measurement effects. Therefore, we could not conduct a thorough Age/ Period/Cohort research design. Nevertheless, the GSS data do afford the opportunity to assess the relationship between religiosity and spirituality and attitudes toward same-sex marriage. Issues related to same-sex marriage are very important at this particular time in American society, and we are able to point to several interesting patterns or trends. We were able to examine the determinants of attitudes toward same-sex marriage within cohorts and compare these effects using national data. Second, this study was also limited to Protestants, Catholics, and no religious preference respondents. This is due to small sample sizes of other religious categories and the analytic strategy to examine each cohort separately for comparisons. The focus was to examine the determinants of attitudes toward same-sex marriage for each cohort. As a result, Jewish respondents, Muslim respondents, and other religious respondents did not have a large enough sample size to be included in the analysis. And third, the analysis is limited to White respondents, African American respondents, and Hispanic respondents. Hence, a number of racial and ethnic groups are not part of the analysis. In addition, we were not able to further examine the variation among Latinos identified by Ellison et al. (2011). Future research should examine comparisons using more diverse data.

Conclusion
This study shows several interesting patterns. First, among the Millennial cohort, respondents who identify with evangelical Protestant denominations show no significant differences from their mainline Protestant counterparts. The bulk of the social scientific literature concludes that White evangelical Protestants are the least likely set of denominations to support same-sex marriage. However, this is not the case for the Millennial cohort. They may think of same-sex marriage in different ways than their older cohorts. Second, attendance at religious services has no effect on Millennials' attitudes. This runs counter to the extant literature for the general population (Sherkat et al., 2011). Attendance at religious services has shown a consistent negative relationship with attitudes toward same-sex relations. That is, increased attendance has been associated with less support for samesex marriage. However, recent research (Walls et al., 2014) found no effect for attendance among a sample of college students when other controls were included in the models. Third, among Millennials, African American Protestants are more likely to support same-sex marriage than their predominantly White, mainline Protestant counterparts. This runs counter to other cohorts where African American Protestants and mainline Protestants are in line with one another. Fourth, while SBNR respondents are more supportive of same-sex marriage for Generation X and the Baby Boomers, this is not the case for the Millennial cohort. If these patterns persist for the Millennial cohort, some aspects of the relationship between religiosity, spirituality, and attitudes toward same-sex marriage are likely to change in the general population.
These results call for additional research on the relationship between religiosity, spirituality, and attitudes toward same-sex marriage for the Millennial birth cohort and Generation X. A smaller percentage of the Millennial cohort identifies as evangelical, but 14% of the sample does express an evangelical religious preference. The results suggest that younger evangelical respondents are more supportive of same-sex marriage than their older counterparts. Perhaps the most striking result is that attendance at religious services has no effect on same-sex attitudes for Millennials. Affiliation and participation are components of institutional effects that are simply not present for this younger cohort. Therefore, going back to "church" later in the life course is not likely to have the same effect as it may have had for the Baby Boom cohort (Newport, 2012).
The subjective aspects of religiosity in the analysis are consistent with previous research. Those who believe in the literal interpretation of the Bible are less supportive of same-sex marriage. The relationship between the literal interpretation of the Bible and attitudes toward same-sex marriage is consistent across all four cohorts. The other subjective component of religiosity is the frequency of prayer. Prayer occurs inside and outside the religious institution. For Millennials and Generation X, as the frequency of prayer increases, support for same-sex marriage decreases. Future research should examine this relationship in more detail.
All indications are that there are changes in the American religious landscape. Younger adults in general are less religious than older adults (Jones et al., 2014;Taylor, 2014). However, our conclusions concerning the Millennial evangelical Protestants are not based on them being less religious. Even though the percentage may be lower than other cohorts, their attitudes do not differ from the mainline Protestants. The results show this is also the case for attendance at religious services. Therefore, our results show a differential impact of religiosity and spirituality on attitudes toward same-sex marriage for the Millennial cohort. As the Millennial cohort and Generation X age, social scientific research is needed to assess the extent to which these younger age cohorts exhibit changes in religiosity, spirituality, and attitudes toward samesex marriage. These two cohorts are the most likely to experience the changes of same-sex marriage in our society. Future research can address a wide range of issues now that law protects fairness and equality in marriage.