Public Attitudes Toward Abortion and LGBTQ Issues

Trends in public attitudes toward abortion, general acceptance of same-sex relationships, and same-sex marriage are examined across regions and parties in the United States. Relative to Americans living outside the South, Southerners are less accepting of same-sex relationships, and these regional differences appear to be increasing over time. On these issues, multivariate analysis suggests that Democrats and Republicans are moving in opposite directions, with Democrats becoming more accepting of same-sex relationships, and Republicans (relatively) less so. With respect to abortion, regional differences do not appear to be increasing or decreasing over time. Both Republicans and Democrats are becoming more accepting of legal abortion, once the effects of other variables have been controlled, but the rate of change is substantially more rapid among Democrats. Thus, party and regional polarization on issues involving sexual morality seems likely to continue.


Article
An enormous body of research has shown that religious values and beliefs have traditionally been strong predictors of certain issue attitudes (for an overview, see Jelen, 2009), and that the moral conservatism of many conservative Protestants (and others) has served as the basis for partisan change and electoral choices. The "God Gap" in party identification (Green, 2007;Killian & Wilcox, 2008;Olson & Green, 2006;Smidt et al., 2010) has been an important source of partisan division for several decades. Although White evangelicals are not monolithically Republican, and many Republican identifiers are not religious or social conservatives, there exists a strong, stable relationship between religious conservatism and attachment to the GOP Republican Party (Patrikios, 2013).
The politics of moral values have contributed to continuity and change in the political distinctiveness of the American South. The high proportion of evangelical Protestants has long rendered Southerners unusually conservative on lifestyle issues. This religio/cultural conservatism has, since at least the 1980s, resulted in a solidly Democratic region becoming disproportionately Republican (Black & Black, 1992;Green, Kellstedt, Smidt, & Guth, 2003) The purpose of this study is to investigate trends in public opinion with respect to two prominent lifestyle issues: abortion and the rights of Lesbians, Gays, Bisexuals, Transsexuals, and "Queers" (LGBTQ). I seek to determine whether, and to what extent, these political questions of "pelvic politics" (Leege & Kellstedt, 1993) continue to animate regional and partisan cleavages in U.S. politics over time.
Of course, at the level of public discourse, LGBTQ rights and abortion continue to occupy prominent places on the public agenda, albeit with quite different trajectories. With respect to the legal rights and public acceptability of homosexuals, public policy is clearly moving in a more "progressive" direction. Even before the Supreme Court's historical decision Obergefell v. Hodges, 1 a number of states had passed laws prohibiting discrimination in housing and in hiring, and several states had either legalized same-sex marriage or have established comparable legal statuses, such as civil unions (National Council of State Legislatures, 2014). Some of these changes occurred as the result of judicial decisions (particularly in the aftermath of United States v. Windsor), 2 but others have come about as the results of actions by elected officials. This, in turn, indicates that public opinion is clearly becoming more accepting of alternative lifestyles (Bowman & O'Keefe, 2004;Jelen, 2011;Wilcox & Norrander, 2002).
By contrast, aggregate public opinion on abortion has been quite stable in the four decades since the Roe v. Wade decision (Jelen & Wilcox, 2003). In recent years, a number of states have enacted legislation, which is intended to restrict access to legal abortion. Such measures include limiting the licensure of abortion clinics, limiting access to facilities in which abortion services are provided, limiting insurance coverage, limiting the use of medication to induce abortion, and providing fetal protection before the attainment of legal viability. A number of antiabortion measures have been introduced in 30 states since 2010 (Eckholm, 2014). It seems plausible to assert that pro-life activists are on the offensive in the United States, and "progressives" on the abortion issue are engaged in a battle to preserve the status quo ante.
This article is intended to compare public attitudes on issues involving reproductive freedom and homosexuality over time, and across regions and categories of party identification. To what extent do attitudes toward abortion and LGBTQ issues reflect regional or partisan differences?

Data and Method
Data for this study were taken from the General Social Surveys (GSS). Three dependent variables are considered. A measure of attitudes toward legal abortion was constructed by computing individual means across seven items.
Respondents were asked whether it should be possible for a woman to obtain a legal abortion if there were a chance for a serious defect in the baby, if the woman's health were endangered by the pregnancy, if the family was poor and could not afford more children, if the woman became pregnant as the result of rape, if the woman was single and did not wish to marry, or if the woman wanted the abortion for any reason. 3 The analysis of the abortion index begins in 1977, and runs through 2012.
Second, a variable measuring general acceptance of homosexuality was computed. To construct this variable, a measure of respondent attitudes toward the morality of homosexuality was dichotomized. 4 Another variable (tolerance toward homosexuals) was computed by taking the mean of items asking whether an "admitted homosexual" should be allowed to give a public speech, to have a book in the library, or to teach at a college or university. 5 The "acceptance of homosexuality" variable was constructed by computing the mean between the morality of homosexuality item and the tolerance toward homosexuals index. 6 The analyses of this index begin in 1973, and run through the 2012 GSS.
A final dependent variable measured specific attitudes toward same-sex marriage, and consisted of a single Likerttype item reading "Homosexual couples should have the right to marry one another." This item was first asked in 1988, and was included on every GSS since 2004.
For all dependent variables, higher scores indicate more conservative responses.
The dependent variables are compared across categories of partisanship, region, and time. To measure partisan identification, dummy variables were computed to tap identification with the Democratic or Republic parties. 7 These dummy variables are employed, rather than the standard seven-point party identifications scale, to allow for the possibility (indeed, the likelihood) that the effects of identification with a particular political party may be asymmetrical. For purposes of the multivariate analyses that follow, "independents" constitute the comparison category.
The regional variable is a dichotomy, separating residents of the Census South (the 11 states of the confederacy, plus Kentucky and Oklahoma) from the rest of the population. 8 To capture the dynamics of the main independent variables, interaction terms between the survey year and region, respondent gender, Democratic identification, and Republican identification are computed.
For purposes of multivariate analysis, the effects of a number of control variables are considered. These include standard demographic variables: age, race (dichotomized into Black and White), 9 and respondent level of formal education. Religious variables include affiliation with an evangelical Protestant denomination, 10 affiliation as Roman Catholic, attitudes toward the authority of the Bible, and attendance at religious services. These items correspond to the conceptualization of religion as consisting of "belonging, believing, and behaving" components (see Green, 2007).
Two additional attitudinal variables are included as controls. One of these is simple ideological self-identification, measured on a seven-point scale ranging from extremely liberal to extremely conservative. A measure of sexual traditionalism is constructed by computing the mean of two items measuring attitudes toward premarital sex and sex between teenagers. 11 Finally, to determine the effects of the variables of interest over time, interaction terms between the survey year and the dummy variables measuring region, partisan identities (Republican and Democrat), and gender are computed. These interaction terms measure the extent to which the effects of the independent variables of interest have changed over time (see Evans & Tilley, 2011, 2012. The analysis strategy is quite simple: For each dependent variable, a multivariate model is computed with the independent variables described above used as predictors. The analysis procedure is ordinary least squares (OLS) regression. To facilitate comparison of the effects of independent variables within and across models, both unstandardized and standardized regression coefficients are reported.

Findings
With respect to attitudes toward abortion, aggregate opinion is quite stable across regions and across time. Southerners are consistently more "pro-life" than non-Southerners, and Democrats are consistently more liberal on abortion than Republicans (data not shown). These findings are quite consistent with previous research (Jelen, 2014;Jelen & Wilcox, 2003;Wilcox & Carr, 2010). Thus, among partisan and regional groups (e.g., Northern Republicans, Southern Democrats), changes in abortion attitudes are nearly nonexistent, and those changes that are observed are statistically insignificant.
By contrast, there has been a trend toward greater acceptance of homosexuality among all regional and partisan groups. Again, Southerners are consistently more conservative than Northerners, and Democrats more accepting than Republicans in both regions. Moreover, the liberalizing trend is slightly stronger among Democrats than Republicans. Among Democrats in both regions, acceptance of homosexuality has increased by approximately 12% of the range of this variable, whereas the corresponding change among Republicans has been about 7%. Thus, simple bivariate comparisons suggest that the parties are diverging with respect to attitudes toward homosexuality, even as identifiers with both parties become more accepting.
Approval of same-sex marriage has also been changing across regions and partisan groups. Since 1988, Republicans acceptance of same-sex marriage has increased by 10% of the variable's range, whereas the corresponding figure for Democrats is 15%. Since 2004, when the issue of marriage equality had become more salient, both Democrats and Republicans increased their approval by about 5% of the range of the variable, although the figure is slightly lower (about 2.5%) for Southern Republicans.
The findings become slightly more complicated when the dependent variables are subjected to multivariate analysis. As Table 1 indicates, there are no major surprises with respect to the effects of demographic and attitudinal variables. Younger voters are significantly more "pro-life" on the abortion question (Wilcox & Carr, 2010), and more accepting of same-sex relationships and marriage (Jelen, 2011). Women are significantly more progressive on these issues, and the interaction terms suggest that gender differences are increasing over time. Race is only related to acceptance of same-sex relationships, with African American respondents more accepting than Whites. As might be expected, more educated respondents are more liberal on all three dependent variables.
All variables measuring religion are significantly related to attitudes toward abortion, same-sex relationships, and same-sex marriage, with the exception that affiliation as a Roman Catholic is not significantly related to attitudes concerning homosexuality. Frequent church attenders, biblical literalists, and evangelical Protestants are all significantly less likely to approve of abortion or any form of same-sex relationship than other respondents. Self-identified conservatives and respondents holding "traditional" attitudes on issues of personal sexual morality are also (predictably) more conservative on questions of abortion and LGBTQ prerogatives.
Of primary interest for this study are the changes in the relationships between the dependent variables, party affiliation, and region. Here, some important differences seem to emerge. With respect to attitudes toward abortion, there is no tendency for Southerners to become more "pro-life" (or "pro-choice") than their Northern counterparts. Moreover, both Republicans and Democrats seem to be adopting more permissive attitudes toward legal abortion over time, although the coefficients for these variables do differ across parties. When the effects of demographic and attitudinal variables are controlled, adherents of both parties appear to be adopting more liberal attitudes on abortion, although the rate of change seems stronger for Democratic identifiers. That is, although respondents who identify with either party are moving in the same (liberal) direction, the parties seem to be diverging at the level of mass publics, because Democrats are changing more rapidly than Republicans. Thus, the abortion issue may continue to be a fertile source of party polarization. By contrast, with respect to general attitudes toward LGBTQ rights, and to the more specific question of samesex marriage, Southerners are becoming more conservative relative to Northerners (Jelen, 2014). Moreover, once the effects of other variables have been controlled, Republican identifiers are becoming more conservative over time, and Democrats more liberal. The coefficients associated with both Party × Year interactions are statistically significant, and the signs of the coefficients associated with party identification have opposite signs for respondents who identify with the Democratic or Republican parties. This suggests that, to the extent that affiliation with a political party is an independent source of political socialization, Republicans and Democrats may be moving in opposite directions with respect to questions of LGBTQ rights.
Of course, it is difficult to assess causality directly with cross-sectional data. Although partisan identification has traditionally been considered the "unmoved mover" in analyses of public opinion, and is usually considered an independent variable in studies of attitudes toward political issues, the policy questions considered here, there exists a strong possibility that issue attitudes may occasion changes in party identification, rather than vice versa. Specifically, Killian and Wilcox (2008) have shown that abortion attitudes are a significant source of individual change in partisanship. More generally, Layman and Carsey (2002) have shown that, in periods of intense party polarization, the causal arrow between party identification and issue attitudes can operate in both directions. That is, although party affiliation can occasion changes in issue attitudes, changing partisanship in response to issue attitudes also occurs empirically. Furthermore, Bishop (2009) has suggested that Americans are self-selecting into politically homogeneous regions, states, or communities. Such ideological or partisan sorting can result in changes in regional differences, even in the absence of individual change.
It is not possible to address such concerns with cross-sectional data. However, a partial test can be constructed, because the GSS contains an item in which the region in which the respondent was raised (specifically, where the respondent lived at the age of 16) was measured. If regional differences could be explained by geographical mobility, the inclusion of this variable in the multivariate models contained in Table 1 should reduce or eliminate the effects of the Region × Year interaction (data not shown). It does not. Unfortunately, the GSS data do not permit a comparable test for changing individual partisanship, and the results presented here should, thus, be interpreted with caution.
It is not clear whether, or to what extent, the observed changes in mass attitudes toward these issues can be attributed to elite socialization. Some analysts (Levendusky, 2009) have suggested that party polarization at the mass level can be explained (at least in part) by polarization among party elites, whereas others (Fiorina, Abrams, & Pope, 2010) are skeptical about the effects of elite socialization. Although this question cannot be answered definitively with the data available for this study, it seems unlikely that the observed partisan and regional socialization can be attributed to popular reaction to the words and behavior of political leaders. The issues in question seem to exemplify what Carmines and Stimson (1980) have termed "easy issues," which are cognitively undemanding, and seem less likely to be susceptible to framing effects. Moreover, the inconsistency of trends among partisan groups across the issues in question makes elite socialization appear an unlikely hypothesis. Despite the clear GOP "pro-life" message on abortion, both Democrats and Republicans seem moving in a "pro-choice" direction, to the extent that the distinctive effects of partisanship can be estimated. The effects of the general measure of acceptance of LGBTQ prerogatives seem approximately equal for adherents of both parties, whereas movement on the same-sex marriage issue seems largely confined to Republicans. An impressionistic description of recent elite-level politics in the United States suggests that partisan cues on issues involving reproductive freedom and unconventional expressions of sexuality are generally consistently progressive or traditional. Mass responses to elite cues would seem likely to move in similar directions among partisan groups.
Conversely, it is perhaps revealing that the opinion movement on the question of same-sex marriage occurs primarily among Republicans. Although Democratic presidents have seemed more sympathetic to LGBTQ rights generally, President William Clinton did not endorse same-sex marriage, and President Obama only did so during his second term. Thus, the hypothesis that partisan cues are an important source of political socialization is consistent with the lack of movement among Democratic identifiers with respect to marriage equality, once the effects of other variables have been considered.

Conclusion
Thus, issues involving aspects of sexual morality, such as abortion, LGBTQ rights, and same-sex marriage, seem likely to remain fertile sources of party polarization, even as mass attitudes on these issues may be becoming more permissive over time. There appears to exist substantial polarization across regions and parties in the United States.
The processes by which these issues are becoming polarized seem quite different, and comparisons across regions and parties reveal distinctions that may be concealed in aggregate analyses. Although aggregate abortion attitudes are quite stable over time, partisan-based change is taking place. Once the effects of other predictors have been taken into account, both Republicans and Democrats are becoming more permissive on the issue of abortion, albeit at different rates. To the extent that party is an independent source of political socialization, these data suggest that partisanship may be a source of short-term polarization, with a long-term tendency toward convergence.
By contrast, attitudes about issues involving LGBTQ rights have, in the aggregate, been changing in a progressive direction. Differences between Southerners and non-Southerners appear to be increasing over time. Moreover, the results presented here suggest that, once distinctions are made across political parties, Democratic and Republican identifiers are moving in opposite directions. Once the effects of religion, ideology, and demographic variables have been controlled, Republicans have become more conservative on issues of same-sex relations, whereas Democrats are becoming more permissive. If these trends continue, convergence on issues involving nontraditional expressions of sexuality seems unlikely.

Acknowledgments
I would like to thank the anonymous reviewers for insightful comments.

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

Funding
The author(s) received no financial support for the research and/or authorship of this article. .889, which does not vary significantly across the years under consideration. 4. Respondents were asked whether sexual relations between members of the same sex were "always wrong," "almost always wrong," "wrong only sometimes," or "hardly ever wrong." This variable was recoded so that "always" and "almost always" wrong were classified in one category, and the other two respondents were combined into a "more accepting" category. Dichotomizing this item in this manner yields the same number of response categories as the tolerance items that are components of the index measuring acceptance of homosexuality. 5. These items exhibit an alpha reliability of .821, which is quite stable over time. 6. Both components of this combined index have a range of 1 to 2. Gamma = .813 overall, and .818 for the 1980s, .793 for the 1990s, and .774 for 2000 to 2010. As might be expected, the components of this index behave similarly to the overall index in bivariate and multivariate analysis. 7. In the analyses presented here, "independent leaners" are coded as partisans. See Petrocik (1974). 8. Specifically, Southern states include Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, and Virginia. Some of the data used in this analysis are derived from Sensitive Data Files of the General Social Surveys (GSS), obtained under special contractual arrangements designed to protect the anonymity of respondents. These data are not available from the author. Persons interested in obtaining GSS Sensitive Data Files should contact the GSS at GSS@NORC.org 9. Of course, it would be desirable to include separate codes for other racial or ethnic groups, such as Asians or Hispanics. However, the values included in the GSS category "other" are too diverse to be useful. 10. This uses value 1 in variable FUND (see Smith, 1990). 11. For each of these items, codes are "always wrong," "almost always wrong," wrong only sometimes," or "hardly ever wrong." Gamma between these items is .658, and is quite stable over time.