KEY Same-sex partnerships Laws in the U.S

Figure 5.3 reports four sets of responses to questions about contemporary gay rights issues from 2004: gay adoption, the legality of gay marriage, a constitutional amendment to define marriage as between a man and a woman (thus constitutionally barring gay marriage), and general attitudes towards gay rights (see Appendix C for exact question wording and poll information). Substantively speaking, these polls, ranging the breadth of contemporary gay rights issues, demonstrate that the American public has polarized on gay rights. All four of the distributions are bimodal or near-bimodal with more mass located in the extreme categories relative to the central categories (the extreme categories are also the two largest frequency categories for each of the issues). Figure 5.4 displays a hypothetical four-category centralized distribution with eighty-percent of the mass of public attitudes located in the two center categories. Figure 5.5 shows the consensus that existed on gay rights attitudes towards offspring being gay in the 1970’s. Over ninety-percent of respondents to the Virginia Slims poll reported that they would find their kid being gay unacceptable.³⁵ The difference between the consensus against homosexual relations in contrast to the near-evenly divided and highly polarized distributions (conflict) in 2004 is plain.

Table 5.4 reports the univariate statistics on all four of the 2004 gay rights questions from Figures 5.4, 5.5, and 5.6. These statistics capture two important aspects of polarization: dispersion (standard deviation, coefficient of variation) and bimodality (kurtosis). Table 5.4 also includes the Virginia Slims question on the acceptability of homosexual relations for ones offspring from 1974, reflecting the consensus position against acceptability that existed at that time. For baseline comparison, I include the hypothetical near-normal four category variable with most of the mass of opinion located in the central categories (80% in center two categories, 20% in extreme categories) from Figure 5.4.

Table 5.4: Dispersion & Bimodality Statistics for Selected Gay Rights Polls & Hypothetical Distribution

Poll	Year	Mean	Standard Deviation	Kurtosis	Coefficient of Variation
Gay Adoption	2004	2.783	1.232	-1.542	44.290
Constitutional Amendment	2004	2.467	1.323	-1.756	53.617
Gay Rights	2004	2.856	1.269	-1.508	44.423
Gay Marriage	2004	2.600	1.312	-1.743	50.492
Gay Adoption (binary)	2004	0.565	0.496	-1.934	87.793
Con. Amendment (binary)	2004	0.299	0.458	-1.223	153.261
Gay Rights (binary)	2004	0.463	0.499	-1.980	107.752
Gay Marriage (binary)	2004	0.482	0.500	-1.997	103.783
Homosexual Relationship	1974	0.979	0.143	43.186	14.591
Hypothetical Centralized Distribution	----	2.5	0.807	-0.482	32.262

Table 5.4 points to the answers to several of the puzzles regarding polarization. It both illustrates the polarized nature of opinion on gay rights today vis-à-vis that in the 1970’s and provides a validity check for the polarization measures used in the aggregate analysis.³⁶ Comparing the standard deviations of the Virginia Slims distribution, that of the hypothetical centralized distribution, and then the four gay rights issues from 2004 (4 cat), the linear increase in dispersion on the attribute is apparent. The standard deviations for the binary, collapsed opinions on gay rights from 2004 (0.500, 0.499, 0.496, and 0.458) are substantially larger than that of the opinion from 1974 (0.143). It is also the case with the coefficient of variation (14.591 for the 1974 poll; 32.262 for the hypothetical CD, and a 48.21 average COV for the 4 polls from 2004). The mean for the opinion on homosexual relationships from 1974 is near 1, indicating the near-consensus of opinion at that time while the means for the opinions from 2004 are near .5 (much more likely to indicate conflict).³⁷

The kurtosis results are consistent with what we know about the included distributions and strong evidence of the increase in bimodality from the 1970’s to the 20o0’s. The four contemporaneous measures of public attitudes towards gay rights (four-cat or binary) all are bimodal, highly conflictual, and polarized distributions both from an absolute sense and in comparison to the opinion on gay rights prior to the onset of the culture war. Both the full and collapsed category measures of gay rights opinion in 2004 are platykurtic and hence indicate bimodality. The Virginia Slims poll on attitudes on homosexual relationships from 1974 reflects the consensus against homosexuals that existed at that time. The standard deviation for the 1974 poll is quite small (0.143) and small relative to the mean, as illustrated in the coefficient of variation (14.591). The kurtosis for this poll is off the charts leptokurtic (43.186), reflective of the single-peaked distribution and the near-total mass of responses located at it (no bimodality). Though the kurtosis measure inaccurately indicates slight bimodality in the centralized distribution, an ordinal comparison of the kurtosis results is consistent with the consensus-to-conflict expectation. It is thus helpful in looking at the relative peakedness differences between the distributions. Furthermore, note that the kurtosis of the distributions for both the two-category and the four-category versions of the polarized polling questions are relatively similar. This tends to weigh against Fiorina’s argument that the Abramowitz transformation (and, by implication, the transformation here) artificially inflates polarization, at least with respect to bimodality.

If we compare kurtosis relative to the other distributions, it illustrates the difference between a consensus position and a centralized distribution. The normal (or near-normal) distribution is not, as some skeptics treat it, the ultimate expression of consensus or moderation. This is tangible evidence reinforcing the theoretical argument I made in Chapter 3: whether or not a shift to a normal distribution on an attribute is indicative of depolarization or polarization is dependent on the shape and dispersion of the prior state of the distribution. In the case of gay rights, the aggregate shift to an on average ‘moderate’ position from the consensus position evident in 1974 is polarization. Aggregate moderation, in this instance, is evidence of increasing conflict (i.e. polarization). As we can see from the distributions in 2004 relative to that in 1974, opinion on gay rights has shifted from one extreme (consensus - unimodality) to the other (conflict - bimodality).

Polarization Trends in Gay Rights Attitudes, GSS 1973-2006

The consistent question-wording of the GSS homosexual relations item permit the direct comparison of polarization models on gay rights. The basic univariate statistics on gay rights from the GSS time-series are reported in Table 5.5. Note the sixth-tenths of a point increase in the mean as well as the three-tenths of a point increase in the standard deviation. Given the variable is scaled from zero to four; these are significant changes in the aggregate public opinion on homosexual relations towards greater polarization on gay rights over the time series.

The ECD measure (extreme category difference: top category minus bottom category on the scale) is the difference between the percentage of respondents that chose the “always wrong” category and those that chose the “nothing wrong at all” with homosexual relations category. This is an important measure with respect to polarization because it provides a direct measure of the changes in the mass at the tails of the opinion distribution on gay rights. Recall, it isn’t enough to have extreme opinions on an issue represented in society in order to generate social conflict. Those opinions have to be held by a sufficient mass of the citizenry in order to provide a venue for political and partisan activism and conflict. If the two extremes have equal mass, then the ECD measure will be zero. The ECD measure shows that there was a large difference between the two theoretical extremes of the gay rights opinion in the 1970’s and into the 1980’s, reflecting the consensus against gay rights that existed at that time. The declining difference between the two extremes tracks the loss of mass at the “anti-gay rights” extreme and the increasing mass at the “pro-gay rights” extreme. This decline in the difference between the two extreme categories is evidence of an increase in dispersion and bimodality. This difference was at its maximum in 1988, right around the point where public opinion on gay rights began to shift. The low water mark (illustrating the equalization trend in the two categories, i.e. conflict) was the 22 percent difference in 2002. There is a quite marked decline in the consensus on gay rights, with a large number of citizens declining to rate homosexual relations as “always wrong” and a corresponding increase in the number of Americans who believe there is nothing wrong with homosexual relations. Also reported in Table 5.5 is the sample N for each year the GSS collected data on the homosexual relations questions. Recall that the statistics for the 1975, 1978, 1983, and 1986 years were interpolated (hence there is no sample N to report).

Table 5.6 reports the polarization regressions of the trends in the mean, standard deviation, coefficient of variation, kurtosis, and extreme category difference models. All indicate a trend towards greater dispersion, and hence increasing conflict on gay rights—increasing polarization in the attitudes of the American public on homosexuals.³⁸ The best model, in terms of goodness-of-fit (.669), is the trend model on the standard deviation in gay rights attitudes. This demonstrates a significant increase in dispersion in opinion on gay rights for the mass public, a key aspect of polarization. The positive coefficient (0.01) indicates an average increase in the standard deviation (i.e. dispersion) of opinion on homosexual relations. The positive coefficient in the mean model (0.019) shows that the average

TABLE 5.6 Trends in Polarization of Attitudes on Homosexual Relations 1973 - 2006

Model: gr = b0 + b1(year) + e	MEAN MODEL	S.D. MODEL	KURTOSIS MODEL	EXTREME CAT DIF MODEL
P.E (S.D.)	0.019*** (0.003)	0.010*** (0.001)	-0.072*** (0.014)	-1.202*** (0.193)
Intercept (S.D.)	-35.017 (5.662)	-18.900 (2.941)	142.393 (27.384)	2439.294 (383.205)
R2	.647	.669	.542	.628
N	25	25	25	25
Pr < |t|	<.0001	<.0001	<.0001	<.0001

* significant at .10 level

** significant at .05 level

***significant at .01 level

response on gay rights in the GSS has moved away from the “always wrong” consensus of the 1970’s and early 1980’s to the contemporary divided opinion on gays today. Again, while this trend is towards the center of the distribution, it is evidence of increasing conflict on gay rights. The ‘anti’ side of the gay rights debate has lost supporters as a percentage of the total population (mass) and the ‘pro’ side of the gay rights debate has gained: drawing the mean on gay rights attitude towards the middle. Both models explain over sixty percent of the variation in the standard deviation and the mean respectively.

The negative coefficient (-0.072) in the Kurtosis model indicates that the distribution of attitudes on gay rights has become more bi-modal over the past three decades. As can be seen in Table 5.5, the distribution goes negative in 1993 and is negative in every survey year after to the present. At the beginning of the series, the kurtosis for the GSS gay rights attitude measure is at a 0.809. By the end of the series, it stands at -1.696. Recall that kurtosis is centered at zero for the normal distribution (mesokurtic). So the distribution on gay rights attitudes flips from leptokurtic to platykurtic over the course of the past three decades. The model shows there has been a significant increase in bimodality in the distribution of opinion on gay rights for the mass public, explaining over half of the variance in kurtosis over the time series relative to the mean kurtosis value (.542). The ECD model shows the declining consensus against gay rights over the time period, with the AW category losing, on average, 1.202 percent of its mass every year of the time-series relative to the NW category. Over half of the variance in the Kurtosis model and sixty-three percent of the variance in the ECD model is explained by the progress of time. Both suggest increasing bimodality on gay rights. Overall, Table 5.6 shows powerful evidence of increasing polarization on gay rights in the American public and is completely consistent with the culture wars thesis. There is increasing dispersion on gay rights (standard deviation model), the extreme categories have become more balanced as opinion has shifted away from extreme opposition to gay rights while extreme support has increased (ECD model), there is increasing bimodality as indicated by the negative coefficient from the kurtosis model, and the mean has shifted away from the anti-gay side of the opinion scale, as we would expect in a shift from consensus located near one pole of the scale towards conflict on the issue (mean model).

Analysis – Partisan Polarization Trends in Public Opinion on Gay Rights

While we have established the loss of consensus on most gay rights issues and on gay rights overall that has occurred over the last three decades, as I noted in Chapter 3, it isn’t mere issue polarization that makes for a culture war. In order for a polarized issue to become politically relevant, it has to be relevant to political competition. The political parties are a prime mover in American political competition. Parties define and shape political conflict, educate and motivate their supporters on issues, and seek to gain supporters by taking advantage of issues that divide the American public. An issue that is polarized and where the parties align along the cleavage of opinion on that issue is the essence of what makes for a cultural war. In the American political system, the threshold for an issue to gain salience among partisans and make its way into the debate over public policy is relatively large given the bias in the electoral system against minority factions (see Durverger’s Law). The GSS item on homosexual rights provides an opportunity to assess the trend in partisan opinion on gay rights and compare the trends in the differences both between the parties and within the parties.

Issues that divide the public and gain political salience present an opportunity for parties to gain supporters by appealing to one side or the other on the issue. Issues on which there is consensus, or which do not lend themselves to policy alternatives, or cross-cut well established partisan cleavages do not lend themselves to partisan and political competition (absent a realignment on the cross-cutting issue). Gay rights falls within the former classification. As a consensus issue in the 1970’s, neither of the major parties had it as a major plank in their party platforms, it was not an issue upon which their candidates campaigned, and consequently the mass public did not identify the parties with a position on gay rights nor did they align themselves with the parties, in whole or in part, based on their attitudes towards gay rights and their perceptions of the parties vis-à-vis gay rights.

That began to change midway through the 1980’s. As the consensus on gay rights died and the emerging social conflict on gay rights emerged, the prospect for partisan competition, partisan alignment on either side of the gay rights debate, and the institution of policy alternatives on gay rights became a possibility. If this story on gay rights is correct, then we would expect that the mass public would increasingly associate the parties with a position on gay rights, and that partisan identifiers would, overtime, increasingly identify with the party’s position on gay rights (either through conversion of partisan identifiers to the pro or anti positions on gay rights or the adoption of a party id by supporters and opponents of gay rights). Thus we would expect to see the party identifiers diverge on their attitudes on gay rights concomitant with the emerging cultural battle over homosexuals and their place in society. This is exactly what we see over this time period, as illustrated in Figures 5.7 - Figures 5.12. The broad classification of party identifiers reflects the emergent partisan polarization on gay rights and this polarization is evident for each of the party identification categories of respondents.

Figure 5.6 illustrates the change in average opinion on homosexual relations among Republican, Democrat, and Independent identifiers in the time series.³⁹ Note that in the beginning of the series in 1973, Republicans, Democrats and Independents are unified in their opinion on whether homosexual relations are wrong. Ninety percent of the respondents, irrespective of party identification, believed that there was something wrong with homosexual relations. The implications for a ‘culture war’ here are clear. In the 1970’s there wasn’t a “culture” war…certainly not on the hot button social issue of gay rights. There was an obvious consensus against homosexuals. No partisan advantage to be had by taking an opposing position and supporting gay rights (indeed, doing so at the time could have rated as political suicide). While some separation occurs over the course of the next decade, there was still relatively little difference (about 5%) between Republicans, Democrats, and Independents on the ‘wrongness’ of homosexual relations. In the early part of the 1990’s, just about the time that Hunter identified the existence of an emerging culture war, Republicans and Democrats begin to increasingly separate on gay rights. As can be seen in Figure 5.7, the difference between Republicans and Democrats tops ten percent and steadily increased through the end of the time series (maxing out at a 20% difference).

This trend in increasing partisan differences is apparent in each of the categories of partisanship on the 7-point scale. As one might expect, the largest disparity exists between strong Republican identifiers and strong Democratic identifiers. In 1973, there was no difference between strong Republicans and strong Democrats on homosexual relations (Figure 5.8). By 2006, forty percent of strong Democrats saw nothing wrong with homosexual relations compared with just fifteen percent of strong Republicans. While the differences are not as marked among weak identifiers (Figure 5.9) or independent leaners (Figure 5.10), there is still an apparent and distinct difference between Republicans and Democrats on gay rights completely consistent with the parties aligning in opposition on attitudes

FIGURE 5.8: GSS – Homosexual Relations are Not Wrong 1973-2006, Strong Republicans and Democrats



FIGURE 5.9: GSS – Homosexual Relations are Not Wrong 1973-2006, Weak Republicans and Democrats



FIGURE 5.10: GSS – Homosexual Relations are Not Wrong 1973-2006, Independent Leaners



FIGURE 5.11: GSS – Homosexual Relations are Not Wrong 1973-2006, Independents

towards homosexuals. The alignment isn’t perfect, however. All three of the parties evidence the overall trend towards increasing tolerance of homosexuals. However, note that Republicans who think there is something wrong with homosexual relations has only declined about 15 percent over the last three decades. The increase in support for gays among Democrats is twice that (Figure 5.6). The behavior of independents is interesting in the later part of the time-series. In the late 1990’s independents tracked closer to the Democrats in their tolerance of homosexuals, but by 2006 they had ticked back to a position just below that of the Republicans. While one can only speculate as to the cause, the prominence of gay marriage as the prime gay rights issue in this decade and the legal interventionism on gay marriage may have produced a backlash against gay rights among Independents.

However, note the distinct difference in the goodness-of-fit measure. The R² for Democrats on the mean trend model indicates that twenty percent more of the variance in the mean on gay rights is explained by the passage of time in comparison to the Republican mean model. The model explains seventy-five percent of the variance for the Democrats but only fifty percent for Republicans. There is a similar distinction between the Republican and Democrat S.D. models, with the Democratic model explaining more of the variance than the Republican model. The answer, I believe, is found in the difference models. The positive coefficient in the R-D mean difference model (0.015) shows that there is a significant increasing difference between Republicans and Democrats on gay rights (R² = .786). While

** significant at .05 level

***significant at .01 level

the difference between S.D. is significant, with Democrats experiencing greater intra-party dispersion than the Republicans, it is a relatively small difference (.002) and a concomitant poorly fit model (R² = .132).

In the 1970’s and the early years of the 1980’s a clear consensus against homosexuals and strong opposition to gay rights existed. As a consequence, the parties largely ignored the issue. The mass public did not differentiate the parties on gay rights, partisan elites had no incentive to stake out positions on gay rights or propose policy alternatives related to the issue, and gay rights was well off the public agenda for all but the small enclaves of gay rights activists. This consensus died sometime in the middle of the 1980’s. Support for gay rights emerged and the same time opposition to gay rights declined. What emerged was an issue ripe for political and partisan conflict. The growing mass of citizens in favor of gay rights encouraged the Democratic Party to woo those citizens (either to convert them to the party or as a consequence of the increasing number of Democratic party identifiers who supported gay rights) by offering policy alternatives such as the protection of homosexuals in the workplace and permitting gays to openly serve in the military. The parties became to be identified with specific positions on gay rights (Republicans opposed; Democrats in support) and gay rights supporters and opponents became an identifiable cleavage between the parties (and thus distinct groups that would need to be satisfied by policies in order to generate money and electoral support). This trend is apparent and significant in all of the partisan measures on gay rights opinion over the past three decades. The parties have clearly diverged on gay rights. In three short decades the Republicans and Democrats in the mass electorate went from having nearly identical viewpoints on gay rights to distinct

Figure 5.12: Which Party is Closest to Respondent on Gay Marriage, 2005⁴⁰

and divergent views on this social issue. It is thus no surprise that gay rights are a significant player in the partisan battles in the culture war from the 1990’s to the present. The American public is divided on gay rights, and hence are divided in their support for either party on gay rights issues (Figure 5.12).

I believe the evidence presented in this chapter paints a very clear picture on political polarization and the culture wars. In the early 1970’s and through the middle part of the 1980’s there existed a strong consensus in the American public that was barely tolerable of and expressly hostile towards homosexuals. In the later part of the 1980’s public attitudes began to undergo a radical transformation in attitudes on gay rights. An increasing number of Americans began to see “nothing wrong” with homosexual relations and were increasingly tolerant of and comfortable with gays. The result of this formation of a second and oppositional camp on gay rights was part and parcel of the emerging culture war in the early part of the 1990’s. Today, whether it be the spectacle of George Takei (Mr. Sulu) getting married to his long-time partner in San Francisco or the heated legal battles over Defense of Marriage acts versus state court-imposed gay marriage, the battle lines are clearly drawn. There is social conflict on gay rights. Here I have shown that a host of gay rights issues have moved from consensus to conflict (with but a few countervailing examples of gay rights attitudes moving from conflict to consensus). Using two different datasets of public attitudes on gay rights, I have shown that there is an overall trend from consensus to conflict on attitudes towards homosexuals, homosexuality, and homosexual rights.

Furthermore, I have shown that the parties have reacted to and aligned along this cultural divide. As public opinion on gay rights diversified, it created an opportunity for the parties to use gay rights to compete for votes. As a consequence the Democrats have aligned much more strongly with the pro-gay rights faction with Republicans sticking with the anti-gay rights faction. The parties have recognized, reacted to, and mobilized on the cultural battle over gay rights. And there is little evidence that this war will end any time in the near future.

This chapter has presented a clear case of increasing political polarization and conflict on one of the hot-button social issues of the day and one of the major fronts in the culture wars. It further serves to demonstrate the application of specific, empirical measures of polarization to test whether opinion on an issue has become more dispersed, more bimodal, and more extreme (e.g. more polarized). However, this chapter on gay rights, while assessing polarization in an important social issue and a key aspect of the culture wars, is narrow in scope. There are other social issues relevant to partisan and political competition and conflict. Furthermore, there are non-social issue dimensions such as economic policy, government spending and regulation, and foreign policy in which polarization (or depolarization) can occur. In the next chapter I expand upon the analysis presented here by employing the measures of polarization to look not just at the range of other social issues that have sparked the culture war, but also at the other major dimensions of political conflict in the United States.