Partisan Ideological Polarization

E_o: No observable trend in the ideological proximity of partisan identifiers.

E_a: A decrease in the distance between partisan identifiers on ideology. (depolarization)

E_a: An increase in the distance between partisan identifiers on ideology. (polarization)

E_o: No observable trend in the proximity of partisan identifiers on the abortion issue.

E_a: A decrease in the distance between partisan identifiers on abortion. (depolarization)

E_a: An increase in the distance between partisan identifiers on abortion. (polarization)

E_o: No observable trend in the party identification of income groups.

E_a: A decrease in the distance between income groups in their party identification. (depolarization)

E_a: An increase in the distance between income groups in their party identification. (polarization)

E_o: No observable trend in the ideological proximity of income groups

E_a: A decrease in the distance between income groups on ideology. (depolarization)

E_a: An increase in the distance between income groups on ideology. (polarization)

E_o: No observable trend in the party identification of religious and secular citizens.

E_a: A decrease in the distance between religious and secular citizens in their party identification. (depolarization)

E_a: An increase in the distance between religious and secular citizens in their party identification. (polarization)

E_o: No observable trend in the ideology of religious and secular citizens.

E_a: A decrease in the distance between religious and secular citizens on ideology. (depolarization)

E_a: An increase in the distance between religious and secular citizens on ideology. (polarization)

As with most of the analyses, I examine the trends in polarization using simple and multiple regression models with the polarization measure as the dependent variable and year (or a transformation of year into time periods) as the independent variable. Significant, positive coefficients indicate a trend towards polarization along the issue dimension in the model while negative coefficients show the opposite.

Equation 4.8: Simple Trend Model of Group Polarization

No apparent trend in polarization may be inferred from the absence of a significant relationship between the group polarization measure and the year variable. Where the party of the presidential administration (ppa) is a theoretically-grounded explanation of group polarization, I include it as an independent predictor of group polarization. This multivariate model is described in Equation 6.2:

Equation 4.9: Multivariate Trend Model of Group Polarization



Analysis & Measures

As I noted in Chapter 3, polarization is characterized by increasing identification with those similar to oneself along some relevant attribute coupled with increasing alienation from those dissimilar to oneself along that same attribute. There are thus three features of polarization: 1) there must be a high degree of homogeneity within each group. 2) There must be a high degree of heterogeneity across groups. 3) There must be a small number of significantly sized groups (isolated small groups or individuals are irrelevant) across the attribute dimension (Esteban and Ray 1994). An empirical measure of group polarization thus incorporates a measure of group size and a measure of the ‘antagonisms’ or distances from all the other groups for a given policy, issue, or ideological dimension. The measure developed here (GP), incorporates both of these key aspects of polarization for the three sets of groups analyzed here: party identification groups, income groups, and religiosity groups.

While the groupings reflected in the party ID, income, and religiosity groups are defendable as valid and reliable constructs of the divisions along partisan, income, and religiosity dimensions, the cut-points established for these groups are somewhat arbitrary. Naturally, if one collapsed categories of these groups it would increase GP overall and, likewise, if one parsed the groups further it would lead to a decrease in GP. However, this concern is mitigated by the fact this analysis examines changes in GP over time. Hence the number of groups remains constant for each survey year. Furthermore, any poor mapping of the groups used onto the actual distribution of mass in the population on these dimensions would work against a finding of polarization, as disparate identifiers on the issue dimension would be included in the wrong groups and thus artificially draw the means of those groupings together. In other words, an inaccurate classification of groups might suggest the absence of polarization when it in fact exists, but would not lead to a finding of a false polarization trend (See Chapter 3 for GP explanation).

Equation 4.10: Empirical Measure of Weighted Group Polarization on Political Dimensions

Where:

This measure of group polarization calculates the total distances between the defined groups on the two dimensions of political conflict that I examine here: ideology and partisanship. It weights the summation of those distances by the sizes of the defined groups, using the ANES sample proportion for the group categories in each survey year. This definition is consistent with the theoretical discussion of polarization (Chapter3), as the maximal polarization in this measure would involve society dividing itself into two groups on the G dimension, with the two groups locating themselves at the extremes of the I dimension. If the groups separate in the I dimension, then that is evidence of a polarization trend. If the groups move towards each other in the I dimension, polarization declines. One asset of this measure is that, even if the number of groups is over-defined (say, we have created foure categories of group but, in terms of their relative location, there really is only three groups) it will not affect the group polarization trends identified in the analysis. Let’s say that Group A and Group B have almost identical positions on ideology. If that is the case, the relative distance between them approaches zero and correspondingly counts little towards the overall polarization coefficient on ideology and would not contribute to an increase or a decrease in the GP measure.

The unweighted group polarization measure simply calculates the squared distances between each of the group on a political dimension irrespective of the size of the group:

Equation 4.11: Empirical Measure of Unweighted Group Polarization on Political Dimensions

Where:

The divisive and charged argument between Justices Kennedy and Scalia illustrate the fact that gay rights and attitudes towards homosexuals are at the cutting edge of contemporary social conflict and is one of the key fronts in the culture wars. However, it wasn’t always so. In the not too distant past, there was a relative consensus in the American public on homosexuals and gay rights. There was a general attitude of distaste for homosexuals and their lifestyles and a bare tolerance for even fundamental civil rights such as the right to hold a job. Homosexuals were not to be trusted in important positions like political office; they were not to be tolerated in jobs that required intimate contact with the public, nor any position which might expose children to gays or their lifestyle. This has changed over the last 35 years. The American public is more accepting of, tolerant of, and positive in its disposition towards homosexuals. This shift in public opinion, however laudable it may be, has also produced a backlash. The loss of consensus on homosexuals and gay rights has produced a cacophony of voices on the one hand demanding a halt to the mainstreaming of homosexuals (if not a complete reversal) with an equally fervent (though smaller in relative size) segment of the population insisting on significant political changes to further normalize homosexuals, homosexuality, and homosexual relations.

Gay rights are a hot button issue in American social conflict. It is squarely on the public agenda in both legislative and legal environments. The parties have mobilized to provide alternative policy choices on gays and gay rights. Activists and political interest groups campaign for and against gay rights and hold officeholders to account on how they do or do not vote on gay rights measures. I will track the course the birth of the culture wars has paved on gay rights— from consensus to conflict. I will delineate the specific gay rights issues that have become more conflictual (or more consensual) over the past several decades. And I will illustrate how this particular front in the culture war has affected the political parties at the mass level and thus spurred the political and policy battles on gay rights we see today.

This relatively narrow focus on just one social issue serves several important purposes in furthering the general analysis of polarization. First, as mentioned earlier, it is a hot-button social issue that is topically relevant to the culture wars thesis and, yet, has received little attention in the culture wars literature. Some scholars have ignored gay rights in favor of social issues such as abortion, while other scholars have suggested that gay rights—while an important social issue and a potential factor in the culture wars discussion—is not currently subject to analysis due to volatility in public opinion and no apparent trends. I remedy that oversight in conducting a thorough examination of public opinion on gay rights and how it has trended over nearly four decades. Second, it allows for a direct examination of the validity and reliability of the empirical measures of polarization developed in Chapter 4. This proof of concept will provide the foundation upon which the later analysis of a number of issues and issue dimensions rests. Third, essentially serving as a case study of polarization, examining just gay rights permits a depth of issue assessment that is not feasible with the broader analysis. The ability to examine gay rights from a number of angles and to consider the variety of aspects to this single social issue adds significant value to the analysis. Thus not only deepening our contextual understanding of this specific social issue, but also sharpening our understanding of polarization in public attitudes on an issue dimension.

In June, 2003, the Supreme Court issued an opinion in Lawrence v. Texas, striking down a Texas sodomy statute and overturning the precedent set in Bowers v. Hardwick, which had declined to find a constitutional right to private, consensual sexual relations a little over a decade prior. While the gay rights movement hailed the decision and culture warriors decried it, one thing is certain irrespective of the constitutional questions: Lawrence was not out of step with the trend in public attitudes on gay rights in America. Gays have steadily gained traction in the public consciousness and attitudes towards gays have shifted significantly in favor of tolerance and normalcy for homosexuals and homosexual relations. One of the prime pieces of evidence against the culture wars thesis is this growing centrism on public attitudes towards homosexuals and gay rights. There is no question that this is happening…and it is happening across demographics and even across party lines. Some culture war skeptics point to this as prima facie evidence against the culture wars thesis. That the aggregate decline in hostility towards gays and the slippage in opposition to gay rights is proof against an emergent divisive social conflict. They are wrong. The growing centrism on gay rights is powerful confirmatory evidence that the culture war has “gone hot.”

The mistake skeptics make in interpreting the data on attitudes towards gays and gay rights is to view centrism as commiserate with political moderation. As I discussed in Chapter 3, while a normal distribution of attitudes may be much more moderate than, say, a bi-polar distribution with the poles shifted to the absolute extremes of an issue dimension, it is not the case when compared to a consensual distribution of opinion with everyone in relative agreement on the issue at hand. In the latter case, a shift towards centrism is, in fact, polarization. It represents an increase in the probability of social conflict and an opening for political activism and mobilization in the policy process. Consensus positions rarely find their way onto the policy agenda, and if they do it is usually as a symbolic or affirmational gesture rather than real, substantive policy. Consensus positions present no incentive for political parties to diverge over related policy alternatives nor do they generate much activism against the consensus position oriented towards persuading electorate-responsive officials (given the small probability of success). When a consensus breaks down, with masses of the public attitude on the issue shifting away from the status quo, the issue becomes ripe for politics. Now money and votes can be procured by appealing to one side or the other of the emergent divide. Policy alternatives can be generated that have some chance of successfully navigating the legislative process. Both sides seek political solutions: either through progressive change or successfully thwarting that effort at change. And as a consequence this issue becomes increasingly relevant to partisanship and electoral politics.

Calm Seas - The Commanding Consensus, 1970 – 1988.

Homosexuality is a taboo in many societies, but it is particularly so in countries with Judeo-Christian traditions. The United States is no exception, with nearly every state having had, at one time, sodomy laws outlawing homosexual relations. While some states had repealed their prohibitions on homosexual sex and many such provisions still on the books went largely unenforced, there was little evidence that the taboo on homosexuality and the societal rejection of homosexuals would ever change, baring extraordinary legal action. Certainly the policy process, subject to popular opinion, seemed an unlikely avenue for change. This is evidenced explicitly in the Virginia Slims American Women’s Poll in 1974 asked 3880 respondents if they would find it acceptable if their grown daughter had a homosexual relationship. 75% of those respondents found it unacceptable with an additional 19% who would accept it but be unhappy about it and would have a strained relationship with that daughter as a result. Only one percent of those respondents reported that they would find the relationship acceptable.³ This is powerful evidence of a strong consensus that perceives homosexuals as deviants, encourages and endorses intolerance towards gays, and opposition to policy innovations on gay rights.

*Harris Survey. Conducted by Roper Organization. June, 1977. Sample N: 1947 respondents.

The halcyon days of consensus on most issues related to gay rights was a mere three decades ago. The 1977 Harris Survey on public attitudes reveals the cut-point from consensus to conflict that existed over what jobs homosexuals could openly hold in society. While the consensus period certainly reflects a shunning of and hostility towards gays, it was not, as some homosexual rights activists would paint it, equivalent to the imposition of Sharia law. Far from public stonings, there was a consensus that homosexuals could hold jobs, make speeches, and participate in civil society. A majority of Americans reported support for equal job opportunities for homosexuals.⁴ While the consensus was accepting of homosexuals in blue collar jobs and positions which require little contact with the going public, there was considerable opposition to homosexuals in positions of authority (congressman, policeman), positions of public trust (social worker), positions which require intimate contact with others (doctor, psychiatrist), religious positions (priest), and positions which involve contact with children (principal, teacher). Though the Harris survey demonstrates that the United States in the 1970’s was far from Afghanistan under the Taliban in its attitude toward homosexuals, it also reveals that homosexuals were perceived as at the fringe of society. Well over half of the public endorsed outright discrimination against homosexuals in all but menial jobs.

The fact Americans were conflicted over the extension of basic civil rights to homosexuals illustrates that a powerful consensus of distaste towards homosexuals was operative. In 1983, Sixty-four percent of respondents said they would not vote for a candidate for president if he were homosexual despite the fact he was well-qualified for the office.⁵ A majority disapproved of homosexual relations between consenting adults for anyone, and less than ten percent found it personally approvable.⁶ Sixty-three percent reported that they were unsympathetic with the homosexual community,⁷ and ninety percent of respondents said they would be upset if their child told them they were gay.⁸ Clearly the American public was more than a little uncomfortable with homosexuals. Against the backdrop of the social consensus against homosexuals, the AIDS epidemic could only have exacerbated fears of homosexuals. While most Americans reported they did not avoid homosexuals because of the AIDS epidemic and were not worried about catching it from them,⁹ majorities reported that they believed discrimination against gays was on the rise as a consequence of AIDS¹⁰ and nearly eighty percent that AIDS, a disease that in the 1980’s was almost exclusively a problem within the homosexual community, posed a significant public health threat.¹¹

Whether a consequence of increasingly favorable representations in the movies and television, the organization of the gay rights movement and the founding of groups like ACT-UP, a reaction to the Supreme Court ruling in Bowers v. Hardwick that upheld the constitutionality of sodomy laws (only a bare majority reported approval), or some combination of exogenous and endogenous shocks to the social system, the tide began to turn in favor of homosexuals and gay rights in the late 1980’s.¹² The consensus against homosexuals was forever shattered. One significant road sign of the radical change that was under way was in the number of respondents who reported they had friends or acquaintances who were gay. In 1985, ninety-one percent of respondents reported they had neither friends nor associates who were gay.¹³ In 1986, seventy-eight percent of respondents reported that they did not have a friend who was gay.¹⁴ By 1992, nearly half of the American public reported knowing someone who was gay or lesbian.¹⁵ A fourth reported having a close, personal friend who was gay.¹⁶ New gay rights issues, such as gay marriage¹⁷ and gays in the military¹⁸ began to emerge during this time period that would become hot topics on the public agenda and embroil a president in controversy early in his tenure. A battle over gay rights in the political arena was brewing.

The cultural conflict over gay rights would come to a head in the 1990’s with the issue of gays in the military. President Clinton, just elected to office and thus ending twelve years of Republican rule in the White House, stumbled into a political briar patch by signaling he would end the ban on gays in the military. President Clinton (and Dick Morris) may have suspected this would be a relatively easy political win, given the consensus that had emerged supporting equal job opportunities for gays. They were wrong. Clinton’s effort quickly became a political firestorm and, consequently, he adopted a Third Way compromise position of “Don’t Ask, Don’t Tell.” While this was intended to end the controversy, the American public was as closely divided on DA,DT as they were gays in the military to begin with (54/47 against DA,DT).¹⁹ The battle over gays in the military was merely a prelude to the war that would be waged, and is still being waged, over gay marriage. In 1996, the United States Congress passed the Defense of Marriage Act, defining marriage as only between couples of the opposite sex at the federal level. The Supreme Court having paved the way with its ruling banning state sodomy laws in Lawrence v. Texas, a number of legislative, judicial, and state referendum efforts have emerged in support of and opposed to gay marriage. Several state supreme courts have ruled that gay marriage is required under their state constitutions, and many state legislatures have moved to enact DOM-like provisions defining marriage as only between a man and a woman. With few exceptions, public opinion on contemporary gay rights issues is passionate and closely divided. While substantial majorities to this day are aligned against gay marriage, those in favor of it are no longer an inconsequential minority. Bare majorities exist in favor of civil unions.²⁰ Nearly half of respondents reported that they have a friend, family member, or colleague who is gay.²¹ And strong majorities (around seventy percent) support extending benefits, social security, and inheritance rights to homosexual couples.²² But the consensus that has developed on benefits and job opportunities are the exception. The rule is close-quarters, evenly divided conflict. The American public is closely divided on whether there should be a constitutional amendment defining marriage as between a man and a woman.²³ Americans are evenly divided on the morality of homosexuality (49/49)²⁴ and gay adoption (40/52 against).²⁵

While a general discussion of the picture of public attitudes on gay rights painted by various and sundry opinion polls over the last 40 years provides a sense of the movement from consensus to conflict—from relative peace to cultural war on gay rights—it does not establish the trend in a rigorous and systematic manner. Are gay rights a burgeoning front in the culture war? Did we have a consensus on homosexuals that has since eroded into two camps—roughly balanced, passionate, and hostile? Has the entry of gay rights in to the continuum of potential vote-getting political issues affected the disposition of the parties on gay rights as well as the policy process? Given these are empirical questions, empirical data is necessary to strike at an answer. I employ two separate data sets to assess the breadth and depth of public opinion on homosexuals and homosexual rights in the United States.

The first set of data is a database of 689 polling questions on attitudes towards gays and gay rights compiled from 1971 to 2007. The poll questions and responses were culled from the IPOLL database developed by the Roper Center for Public Opinion Research.²⁶ The observations for the Gay Rights Database (GRD) were selected from the IPOLL database using a search function that looked for polling questions with the word “gay” or “homosexual” in them. This search found 1523 polling questions that met the selection criteria. From there I chose polling questions that reflected public attitudes towards gays, homosexuals, gay rights, association with gays, gay marriage, gays in the military, etc.

Excluded from the data set were fact-based questions, such as “On a scale of one to seven, how would you rate your own personal risk of contracting 'AIDS'?”²⁷ Also excluded where questions that asked the respondent about the future, such as “In the future, do you think 'AIDS' (Acquired Immune Deficiency Syndrome) will affect more people who are homosexual, or gay, or do you think it will affect more people who are heterosexual, or straight?”²⁸ Some questions asked the respondent to rate one of the political parties, government, or politicians have acted on gay rights. Since it was not clear from the question whether the respondent thought the performance on gay rights issues was good or bad, those questions were excluded as well. In short, if a question was determined to be unrelated to public attitudes on gay rights, or what that attitude is was not clear, the question was excluded. In order to keep a relatively even distribution across questions and to avoid repetitive entries, some attitude-related questions were excluded when a sufficient number of questions on that attitude had been obtained for that year.

The response categories for each question were collapsed into a binary set: Pro-Gay and Anti-Gay categories. This was done to facilitate cross-poll comparisons. The minimum number of questions in the data base for each year is one polling question, the maximum number is 76 (2004) with an average of 21.5 questions per year.²⁹ There are polling questions from 103 different polls in the database.³⁰ There are a number of polls from which there is just one question included in the GRD. The most polling questions (74) in the database are from the PSRA / Newsweek Poll. The average number of respondents per poll in the GRD is 1270.69 with a minimum of 549 respondents (ABC News / Washington Post Poll, January 1993) and a maximum of 8769 respondents (CBS News Exit Poll, November 1978).

The second data set consists of counts and statistics generated on the homosexual relations question asked consistently since 1973 in the General Social Survey.³¹ The exact question wording and response frequencies for the GSS attitude on homosexual relations item is reported in Appendix F.

Both the GRD and the GSS time-series data have advantages and disadvantages in examining trends in public attitudes toward homosexuals and gay rights. The GRD covers a larger variety of gay rights issues, and the time series of GRD polling questions permits an analysis of the specific gay rights issues contemporaneous to when they entered the public agenda and to track them over the period that they were a hot public issue. The GRD allows for a more nuanced assessment of attitudes towards gay rights and thus a more valid assessment of the continuum of gay rights public opinion. It also provides for a variety of different question wordings on the same issue, thus decreasing the chance that question wording overstates or understates gay rights attitudes for the entire sample. But this is a double-edged sword, as the different formulations on a particular issue can influence respondent answers and hence muddy the waters. Differences in responses may be a consequence of different question wording and not an actual change in public opinion on gay rights. While the GSS data cannot differentiate between opinion on gay marriage and civil unions, or gays in the military versus social security benefits for gay couples, it does provide a consistent question wording across the time series. This makes it potentially a more reliable measure of public opinion on gay rights, but it is restricted to just the single aspect of gay rights covered by the GSS question on homosexual relations.

The two primary source variables for the analysis are the Pro-Gay / Anti-Gay collapsed (when necessary) binary variable from the GRD, and the attitudes towards homosexual relations variable from the GSS. As is evident from the breakdown of the GSS variable in Appendix F, the survey years 1975, 1978, 1983, and 1986 did not ask the “attitudes towards homosexual relations” question. I interpolate those data points for the GSS analyses using the two most proximate data points (survey year prior, survey year after) and the simple linear interpolation method.³² This is done for easy of graphing, as exclusion of the interpolated data points does not affect the substantive results of the analyses.³³

The primary classification variable for the GRD is the issue type variable, used to subset the data set by specific gay rights issues. Appendix D has a description of the kinds of questions included in each of the categories of the issue type variable. I will briefly review the categories used in this analysis. In the time-series conflict/consensus trend analysis, the operative rule was that a category must have a minimum of 14 poll questions to be included. The self-explanatory categories in the issue type variable are: gay marriage, gay adoption, morality of homosexual relations, the gay rights movement, civil unions, constitutional amendment, comfort, teach, inheritance, and benefits. All of those categories include questions on those clear gay rights topics (see Appendix D for further clarification). A bit less clear are the following: The “vote” category includes questions that ask the respondent whether they would vote for a candidate they knew to be homosexual. The “candidate” category includes questions where respondents indicate that they think whether or not the candidate for office is important, whether it would influence their vote, and if the public has a right to know that information. The “friends” category includes questions where the public reports on whether they have friends, associates, colleagues, or relatives who are gay. The “jobs” category references questions on homosexual job discrimination and equal opportunity rights in jobs for homosexuals. The “military” category refers to questions both about permitting openly gay soldiers to serve and the “Don’t Ask, Don’t Tell” policy.

The year variable is self-explanatory. It was re-coded for the regression analyses (ranging from 1 to 32 for the GDRB and 1-25 for the GSS) for ease in reporting the intercept parameter estimates in the tables. The party identification variable used for the GSS analyses is the standard ‘PARTYID’ variable used in their semi-annual surveys. The GSS uses a slightly different set of questions to produce the standard 7-point party identification scale. The GSS classifies partisan identifiers as either “strong” or “not very strong” Democrats or Republicans. The independent / leaner category asks respondents which of the two parties they are “closer” to. Despite the different question wording, the scaling of the GSS party identification variable is identical to that of the party ID variable in the NES (0-6).

Simple statistics for the gay rights analysis are reported in Table 5.1. The most frequent category in the GRD data set is gay marriage, largely a function of the number of polls included from 2004 and the fact that the number of polls and amount of polling has steadily increased over the last few decades. Also reported in Table 5.1 are the mean Anti-Gay attitudes in the GRD for the entire data set. The most consensual anti-gay rights attitude for the entire data set is the attitude towards the prospect of one’s child deciding he or she is gay. The most consensual pro-gay rights attitude is on gays having equal job opportunities and opposed to discrimination against gays in the workplace. The most contentious gay rights issues are civil unions and whether there should be a constitutional amendment to define marriage as only between a man and a woman. It is not a coincidence that these are also two of the more recent gay rights issues to enter the public agenda, both falling within the “culture wars” time horizon (1993 and 2003).

The only significant exceptions to the ‘recent = conflict’ rule are gay marriage and inheritance rights. Gay marriage has been consistently opposed by a strong majority of citizens since it was introduced to the policy environment. Inheritance rights have likewise enjoyed majority support since pollsters in the GRD began asking respondents to rate their opinion on gay rights to inheritance. If you take 1990 as a cut point and look at the issue categories with 10 or more polling questions, only inheritance falls outside the 43 – 52 range (Table 5.1). While a consensus may have emerged on granting homosexuals inheritance rights, the same cannot be said about gay marriage. When coupled with civil unions, gay marriage is one of the more contentious contemporary gay rights issues. Fourteen states have adopted gay marriage or civil union statutes (Figure 5.2). Furthermore, in response to the gay marriage ‘threat,’ a number of states have amended their constitutions to forbid gay marriage.

Though gay marriage is a loser on the national stage (not unsurprisingly, both 2008 presidential candidates were against it), the strong minority on gay marriage has significant political influence in the Northeast and the Northwest. In contrast, gay marriage is strongly opposed in the Southern and Western states (otherwise known as fly-over country). As we saw with Lawrence v. Texas as well as the

* http://en.wikipedia.org/wiki/File:Samesex_marriage_in_USA.svg