Changing Public Opinion on Same-Sex Marriage: The Case of California

Abstract

Though public opposition to same-sex marriage seems reasonably stable nationally, support in California has grown substantially in the past two decades. Using data from six Field Polls of Californians since 1985, we explore the roots of that growth in individual attitude change and population changes. Cohort replacement can explain half the growth. Although all groups of Californians say that they have become more accepting of homosexual relations since they turned 18, the pattern is strongest for liberals, Democrats, and the less religious. These groups have also become much more supportive of same-sex marriage, while conservatives, Republicans, Protestants, and African-Americans appear at least as opposed today as they were two decades ago.

Explaining Rising Support for Same-Sex Marriage in California

San Diego's conservative Republican mayor recently held a highly emotional press conference to announce that he had changed his mind about same-sex marriage. Despite his previous opposition, he had decided that civil unions were a separate but unequal alternative to marriage, and that he could not look his lesbian daughter or gay and lesbian staff members ‘in the face and tell them that their relationships—their very lives—were any less meaningful than the marriage that I share with my wife’ (Vigil 2007). While rare in such a public setting, similar changes in opinions have been taking place across California. Opposition to same-sex marriage remains strong and reasonably stable nationally, but Californians have become strikingly more sympathetic, are now nearly evenly split on same-sex marriage, and have barely reacted as their state legislature has acted to recognize same-sex relationships.

To understand this shift in public opinion, we look at both individual mind-changing and demographic changes. Most studies on public support for gay rights consider either general trends in opinions or demographic characteristics that predict more or less accepting attitudes. Few look at which groups are changing their minds or whether demographic predictors have altered. We take advantage of California's anomalous opinion shift and some unique data on Californians' attitudes to study change in two ways. Using a 2006 poll, we examine which individual characteristics predict people's own assessments of how their attitudes toward homosexuality have changed. Then, combining data from six polls since 1985, we examine which characteristics predict the most rapid gains in support for same-sex marriage.

To show the importance of understanding attitude change on homosexual relations and on same-sex marriage, we begin by describing briefly the history of one of the defining ‘morality policy’ issues of our era, and the likely role of public opinion in explaining the widely divergent paths the states have taken. We then summarize the research on correlates of attitudes toward homosexuality and gay rights and on trends in public opinion before focusing on attitude change in California. We reach several conclusions. First, cohort replacement explains just over half of the rise in support for same-sex marriage, leaving only half to be explained by mind-changing or other factors. Second, several variables that strongly affect acceptance of homosexuality do not predict changes in attitudes. For instance, older, male Californians are much more likely than younger, female ones to condemn homosexual relations and oppose marriage rights, but they are nearly as likely to report having become more accepting since they were 18, and the level of support for same-sex marriage in both groups has risen at a comparable rate. Third, some variables that influence support for same-sex marriage today did not matter nearly as much 20 years ago. Compared to 1985, Californians' opinions on homosexuality and gay rights issues today are much more strongly split along ideological, partisan, and religious lines. Liberals, Democrats, and the less religious have become more supportive of same-sex marriage, but conservatives, Republicans, and Protestants have not. Similarly, African-Americans' support for same-sex marriage has not risen, while that of non-Hispanic whites, Latinos, and Asian-Americans has.

Public Opinion and Same-Sex Marriage

Despite long debate within the gay and lesbian community (Cole 1972; Egan and Sherrill 2005), same-sex marriage was so unthinkable to most heterosexuals that it did not become an issue until the Supreme Court of Hawaii ruled that denying same-sex couples the right to marry may have violated that state's constitutional prohibition on sex discrimination (Baehr v. Lewin 1993). Once the issue appeared on the national agenda, however, opposition was so intense that Congress passed the Defense of Marriage Act (DOMA) within months, two dozen states passed ‘mini-DOMAs’ within two years, and another dozen have done so since (Cahill 2004). When Massachusetts' high court ordered that state to issue marriage licenses to same-sex couples (Goodridge v. Department of Public Health 2003), President Bush argued for an amendment to the U.S. Constitution to prohibit same-sex marriages, and 23 states amended their own constitutions to do so.

Nonetheless, several states have created legal protections for same-sex couples, typically under pressure from their courts. The Massachusetts state legislature refused to amend the state constitution to overturn the decision made by its supreme judicial court. Vermont, Connecticut, New Jersey, and New Hampshire now recognize civil unions, Hawaii has ‘reciprocal beneficiaries,’ and Oregon has domestic partnerships. The California state legislature may have gone the furthest without judicial pressure. It created domestic partnerships in 1999 and has expanded their legal benefits in each subsequent legislative session, until same-sex couples have the vast majority of marital rights and responsibilities the state can grant. The legislature has twice passed civil marriage equality—the only legislature to have done so—although Governor Schwarzenegger has vetoed it both times (Equality California 2007). Despite passing an initiative in 2000 declaring ‘[o]nly marriage between a man and a woman is valid or recognized in California,’ California voters have not reacted strongly to the legislature's actions—an initiative to amend the state constitution failed to gather enough signatures, same-sex marriage was not an issue in the 2006 campaign, and no legislator who supported civil marriage for same-sex couples was voted out of office.

How can legislatures in California and a few other states be recognizing same-sex relationships when most states are taking such strong steps to deny marriage equality? Interest-group strength clearly plays a role; Haider-Markel (2000, 298) rates California's gay interest-group resources as the second highest in the country.¹ California also has several gay and lesbian legislators who have taken key roles in moving domestic partnership and marriage legislation forward (on the role of gay and lesbian elected officials see Haider-Markel, Joslyn, and Kniss, 2000). However, state policies also generally reflect the ideologies and beliefs of the state's citizens (e.g., Erickson et al. 1993; Norrander 2000; Wright, Erickson, and McIver 1985). The connection is especially strong on simple, politically salient issues where legislative action is highly visible (Burstein, Bauldry, and Froese 2001; Carmines and Stimson 1980; Gormley 1986; Haider-Markel and Meier 1996). Gay rights issues are typically simple, salient, and visible (Mooney and Schuldt 2006), and Lewis and Oh (2008) find a strong link between public opinion and policy: the eleven states with the most supportive public opinion have taken public action to recognize same-sex relationships, and ten of the twelve states with the most hostile public opinion have passed both legislative and constitutional bans.

Understanding public opinion is therefore important for understanding the future of same-sex marriage in the United States. Assessing national trends in support for same-sex marriage is hampered by inconsistent question wording and by the relatively recent appearance of surveys on the subject, given that the first question on a national survey was in 1988 (Bowman and Foster 2006). Based on this imperfect evidence, opinion appears reasonably stable nationally—55 to 65 percent oppose it and only 30 to 35 percent favor it (Brewer and Wilcox 2005; Rogers 1998; Yang 1997). In analyzing the survey data, Brewer and Wilcox (2005, 600) conclude that ‘from the early 1990s to the present . . . there is no sign of a dramatic trend toward greater support.’Lewis and Oh (forthcoming, 2008) find a weak upward trend, but only when they ignore downturns in high-controversy years (1996 and 2004). Bowman and Foster (2006, 21-3) present several series of polls that suggest erratic changes (see Figure 1). The CNN/Time series shows growth in the early 1990s, but then drops in the wake of discussion on DOMA in 1996. The Pew Research Center and Gallup series show early growth, possibly because they start in 1996, but both drop in 2003-04 when same-sex marriage is back in the news.

Figure 1. Trends in Support for Same-Sex Marriage

In contrast, the Field Poll shows steadily increasing support in California on the same question asked seven times since 1977, from 28 percent in 1977 to 30 percent in 1985, 37 percent in 1997, and 43 percent in four polls since 2003, with no drops when controversy peaked. This particular question has never been asked of a national sample, making cross-state comparisons difficult, but Lewis and Oh (2008) estimate support in California to be higher than in all but seven states.² Thus, though support for same-sex marriage remains a minority position in California, it is higher and shows clearer signs of growth than in most parts of the country.

Predictors of Differences in Attitudes

Why has support grown in California? The extensive empirical research on individual attitudes toward homosexuality and gay rights focuses on predicting differences rather than changes,³ but it suggests several conclusions that may also hold for changes. First, younger people tend to be more accepting. Although people may become more sexually conservative as they age, changing socialization patterns for different age cohorts is the more common explanation; as society has become more accepting of gay people and the media has presented increasingly positive images of gay men and lesbians, succeeding generations are raised with more tolerant attitudes (Becker 2006; Davis 1992; Fiorina 2005; Levina, Waldo, and Fitzgerald 2000; Mayer 1992; Tropiano 2002).

Second, acceptance tends to rise with education, which exposes one to more diverse perspectives, teaches social and political tolerance (Bobo and Licari 1989), and increases the likelihood of knowing openly gay people (Herek 1988), although the effect of education is surprisingly weak on opinion about same-sex marriage (Lewis 2005). Third, more religious people—those for whom religion, prayer, and attendance at religious services are more important—tend to be more condemning of homosexuality, especially if they belong to more fundamentalist Protestant denominations (Finlay and Walther 2003; Herek 1988;Jelen 1993; Laythe et al. 2002; Olson, Cadge, and Harrison 2006; Pew Research Center 2006). Fourth, liberal Democrats tend to be more accepting than conservative Republicans, especially as the “family values” agenda of social conservatism has become a more dominant aspect of political conservatism (Jelen 1993; Kaufmann 2002). Fifth, women tend to be more accepting, perhaps because men find homosexuality more threatening (Davies 2004; Franklin 2000; Herek 1998, 2000; Kite and Whitley 1996; Ratcliff et al. 2006; Whitley 2001). Sixth, blacks and “Asians” (i.e., persons of Indian and Pakistani origin living in Britain) tend to be less accepting of homosexuality than whites, although not less supportive of government action to end anti-gay discrimination (Davies 2004; Lewis 2003; Pew Research Center 2006); Latinos and non-Hispanic whites do not appear to differ significantly (Bonilla and Porter 1990, cited in Herek and Gonzalez-Rivera 2006). Seventh, knowing someone lesbian or gay has a positive impact on acceptance of homosexuality and support for gay rights (Herek and Allen 2007; Lewis 2007; Mohipp and Morry 2004; Pettigrew and Tropp 2006).

Analyses of changes in attitudes toward homosexuality and gay rights over time tend to focus on trends (e.g., Bowman and Foster 2006; Rogers 1998; Yang 1997) and responses to particular events (e.g., Stoutenborough, Haider-Markel, and Allen 2006). They sometimes assess the impact of cohort replacement (e.g., Rogers 1998; Yang 1997), but only a few studies devote much attention to subgroups. Loftus (2001) finds that acceptance of homosexuality rose more rapidly for women than men and for whites than blacks in the 1990s, and Stoutenborough and others (2006) find women and blacks responding more strongly, albeit in different directions, to Supreme Court decisions in gay rights cases. Lindaman and Haider-Markel (2002) find little evidence of a growing partisan divide on gay rights at the mass level, despite striking disparities between Republican and Democratic elites. Likewise, Fiorina (2005) finds only moderate differences between Democrats and Republicans in the electorate, with both becoming more accepting. In contrast, Persily, Egan, and Wallsten (2006) find that political ideology is an increasingly powerful predictor of support for same-sex marriage, with support rising much faster for liberals than conservatives, and note that support among blacks has dropped substantially. In sum, the effects of age, education, religion, ideology, party identification, sex, and race on attitude differences are fairly clear, though it is not as clear whether those effects are growing or shrinking.

Although individual attitude change contributes to collective attitude change, public opinion can change dramatically over time without a single person changing his or her mind, simply due to changes in the population (Mayer 1992). The most important source of change on many issues is cohort replacement, as older people die and children mature (Davis 1992; Mayer 1992). Rogers (1998) and Yang (2003) attribute much of the rising support for gay rights over the last quarter of the twentieth century to cohort replacement. Migration patterns and different birth and death rates may also shift public opinion by changing the racial, ethnic, and religious mix of the population.

Data and Methods

Self-Perceived Attitude Change

The February 2006 Field Poll asked a simplified version of a standard General Social Survey (GSS) question: ‘Do you think sexual relations between two adults of the same sex is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?’ It followed up with: ‘Looking back, have your opinions about homosexual relations between consenting adults changed a lot, a little, or not at all from the opinions you held when you were 18 years old?’ It then asked those whose opinions had changed whether they had become more or less accepting. For our dependent variable, we simplify respondents' perceptions of their own opinion change to ‘less accepting,’‘no change,’ and ‘more accepting,’ ignoring the extent of the change.

We began with contingency table analyses of attitude change. Our independent variables included age, education, religion, party identification,⁴ political ideology, gender, and race/ethnicity. Due to unique questions in the 2006 survey, we were also able to divide respondents into those with close gay friends, others with gay family members or colleagues, and those who did not personally know anyone gay; to distinguish “born-again” from other Christians, and to use respondents' rating of the importance of religion in their own lives on a four-point scale from ‘not very’ to ‘extremely’ important.

We followed with logit analyses for (1) attitudes toward same-sex sexual relations; and (2) reported changes in those attitudes. We performed ordered logit analysis on a three-level ‘acceptance of homosexuality’ variable, combining ‘almost always wrong’ and ‘wrong only sometimes’ into a single value to meet the parallel odds assumption. As the opinion change measure did not meet the parallel odds assumption,⁵ we reported separate binary logits for whether respondents had become more or less accepting of homosexuality.

Same-Sex Marriage

We obtained individual-level data for six of the seven Field Polls since 1977 that included the question: ‘Would you approve or disapprove of a law that would permit homosexual people to marry members of their own sex and to have the regular marriage laws apply to them?’ We dichotomized the responses into ‘approve’ and other responses, the latter value combining the ‘don't know’ and ‘disapprove’ responses. The contingency table analysis examined support for same-sex marriage at three times: 1985, 1997, and 2003-06. The independent variables matched those in the earlier analyses of perceived attitude change, but we did not include the questions about religious conservatism, religious intensity, and personal acquaintance with lesbians and gay men, as they were not asked in the earlier surveys. We treated ‘age’ as the decade when a respondent was born, rather than the age in years, to make it possible to track attitude change over time.

We performed three versions of the logit analysis. The first included only a set of dummy variables for decade of birth and survey period (a variable coded 1 for 1985, 2 for 1997, and 3 for 2003-06). The survey period coefficient estimated the amount of individual attitude change after subtracting the effect of cohort replacement. The second added the other demographic variables,⁶ controlling for the effects of other changes in the population, but assuming that those effects remained constant over the period. The third added interactions between survey period and the other independent variables to determine whether the effects of these variables have changed over time.

In discussing the findings from the logit analyses, we translated logit coefficients into expected changes in probabilities holding the other independent variables at their means. Logit coefficients represent changes in log-odds—the natural logarithm of the odds, which is the ratio of the probability of, for instance, supporting same-sex marriage to the probability of not supporting it [P/(1 − P)]—rather than the simpler changes in probabilities. However, logit analysis assumes that the independent variables have a linear impact on the log-odds, which means that they have a nonlinear impact on probabilities. Logit coefficients therefore translate into different probability changes depending on the values of all the independent variables. The choice of holding them all at their means is somewhat arbitrary, but no clearly superior solution exists.⁷

Findings

Self-Perceived Change

Slightly over half the respondents in the 2006 survey say that their ‘opinions about homosexual relations between consenting adults’ have changed since they were 18, and about two-thirds of those respondents have changed ‘a lot.’ Those who have become ‘more accepting’ outnumber those who had become ‘less accepting’ by 5 to 1, producing a net change of 36 percent becoming more accepting (45 percent minus 9 percent; Table 1). In the following discussion, we focus on this ‘net change’ measure, the difference between the percentages reporting more and less accepting attitudes.

Table 1.  Attitude Change Toward Homosexuality since Adulthood, 2006

	Net Change^a	% Less Accepting	% No Change	% More Accepting	N
^a  Net Change = % more accepting − % less accepting.
Overall	36	9	47	45	900
Acquaintance with gay people
Gay friend	46	4	46	50	309
Gay family member or acquaintance	34	7	52	41	261
None	28	15	42	43	329
Party identification
Democrat	44	5	45	49	423
Independent	32	11	46	43	163
Republican	26	13	49	39	313
Ideology
Liberal	43	5	48	48	219
Moderate	32	11	46	43	163
Conservative	27	11	50	38	265
Importance of religion in life
Not very important	46	3	47	49	179
Fairly important	38	7	48	45	206
Very important	39	10	41	49	217
Extremely important	25	13	49	38	297
Religion
Born-again Christian	27	15	43	42	255
Other Protestant	41	4	51	45	149
Other Catholic	38	8	46	46	215
Jewish	41	4	51	45	23
Other religion	33	11	46	44	117
No religion	44	3	50	47	140
Race/ethnicity
Non-Hispanic white	34	6	54	40	465
Latino	41	13	33	54	256
Black	31	6	58	37	47
Asian	31	14	41	45	116
Age
Under 40	40	9	43	49	358
40-54	34	10	46	44	295
55 and over	32	8	53	40	248
Level of education
High school or less	33	15	36	48	323
Some college	34	8	50	42	225
College graduate	39	4	54	43	349
Gender
Female	37	7	49	44	465
Male	36	10	44	46	435

Every group has a net change toward more accepting attitudes, but liberals, Democrats, those with gay friends, and the less religious have the biggest net changes. Net change toward acceptance is 18 percentage points higher for respondents with gay or lesbian friends than for those who did not know any lesbians or gay men (46 percent minus 28 percent). Net change is also 18 points larger for Democrats than Republicans, and 16 points higher for liberals than conservatives. Those for whom religion is ‘not very important’ are 11 points more likely to have become more accepting, and 10 points less likely to have become less accepting than those for whom religion was ‘extremely important,’ for a larger net gain of 21 points. Born-again Christians stand out among the religious groups for having the smallest net change.

Other variables matter less. Protestants who have not been born again, Catholics, Jews, and the nonreligious have comparable net increases in acceptance. Although less time had passed since their 18th birthdays, respondents under 40 show more net change toward acceptance than their elders, though only about 8 points more than those 55 and over. Education also has some positive impact on attitude change, though net change for college graduates is only 6 points higher than for those with no college degrees. Latinos show more net change toward acceptance than non-Hispanic whites, blacks, and Asians, but a high percentage of Latinos have also become less accepting. The patterns of attitude change for men and women are very similar.

Logit models (Table 2) suggest that only a few of the variables that affect attitudes toward homosexual relations also influence attitude change. The model for the determinants of acceptance of same-sex sexual relations (column 1) holds few surprises. Compared with those who say religion is ‘extremely important’ in their lives, those who say it is ‘not very important’ are predicted to be 43 percentage points more likely to also say that homosexual relations are ‘not wrong at all’ (holding the other variables at their means, as in all these comparisons). Liberals are 34 percentage points more likely than conservatives to respond ‘not wrong at all.’ The probability of calling homosexual relations ‘not wrong at all’ is 16 points higher for women than comparable men, 42 points higher for Jews and 16 points higher for Catholics than mainstream Protestants, 5 points higher for 40- than 50-year-olds, and 8 points higher for college than high school graduates. Respondents with close gay friends and those who know lesbians and gay men in other ways are 32 points and 8 points more likely than those who do not know anyone gay to say ‘not wrong at all.’ Party identification and race do not have significant effects.

Table 2.  Logit Models for Attitudes toward Homosexual Relations

	Acceptance of Homosexuality	Became more Accepting	Became Less Accepting
Notes: Robust z-statistics in parentheses. Coefficients significant at * 10%,** 5%,*** 1%.
Has close gay friend	.911***	.522**	−.899**
	(3.33)	(2.44)	(2.37)
Has gay acquaintance or relative, but not close friend	−.014	.181	−.528
	(.06)	(.88)	(1.54)
Conservatism (1-3)	−.674***	−.052	.183
	(4.35)	(.43)	(.86)
Republicanism (1-3)	−.194	−.165*	.418**
	(1.64)	(1.70)	(2.21)
Male	−.654***	.137	.461*
	(3.43)	(.85)	(1.68)
Education (years)	.091***	−.041	−.152***
	(2.98)	(1.56)	(3.31)
Age (years)	−.022***	−.002	−.004
	(3.03)	(.40)	(.48)
Importance of religion in life (1-4)	−.646***	−.184**	.131
	(6.34)	(2.10)	(.86)
Born-again Christian	−.351	.147	.576*
	(1.56)	(.71)	(1.79)
Catholic	.607**	.134	−.200
	(2.45)	(.62)	(.60)
Jewish	1.961**	.062	.395
	(2.53)	(.15)	(.35)
Other non-Christian religion	.421	.052	.735
	(1.48)	(.18)	(1.56)
No religious preference	.266	.027	−.358
	(.76)	(.10)	(.51)
African-American	−.478	−.104	−.356
	(1.12)	(.29)	(.55)
Latino	.046	.557**	.009
	(.16)	(2.46)	(.02)
Asian-American	−.257	.258	.903**
	(.67)	(.89)	(2.21)
McFadden's pseudo-R^2	.21	.04	.14
Observations	858	898	898

When looking at the logit model for attitude change rather than attitude content, only four variables have a clearly significant impact. Holding the other variables at their means, respondents with close gay friends are 13 percentage points more likely to have become more accepting and 5 points less likely to have become less accepting than those who do not know anyone gay. Democrats are 8 percentage points more likely to have become more accepting and 4 points less likely to have become less accepting than Republicans.⁸ Those who say religion is ‘extremely important’ in their lives are 13 percentage points less likely than those who say it is ‘not very important’ to have become more accepting; they are an insignificant 2 points more likely to have become less accepting. Consistent with the analysis in Table 1, Latinos are 14 points more likely than comparable non-Hispanic whites to have become more accepting, although, overall, they are no more likely to be accepting of homosexuality than non-Hispanic whites.

Two other variables have significant coefficients in one of two models, but they appear to affect the probability, rather than the direction, of attitude change. College graduates are 4 percentage points less likely than high school graduates to have become less accepting, but also 4 points less likely to have become more accepting, although the latter effect falls short of statistical significance. Asian-Americans are 6 points more likely than whites to have become less accepting, but also a statistically insignificant 5 points more likely to have become more accepting. Age, political ideology, and gender have important impacts on acceptance, but not on changes in acceptance.

Support for Same-Sex Marriage

Support has risen 13 percentage points across three survey periods: 1985, 1997, and 2003-06 (Table 3). We attribute 6 points of that rise to net change in individual attitudes, as support has risen only 5 to 8 points within each birth cohort. A logit model (not shown) on all survey years combined, with ‘survey period’ and a set of dummy variables for decade of birth as the only independent variables, confirms that respondents in the 2003-06 surveys are 6 points more likely to favor same-sex marriage than respondents from the same birth cohort in the 1985 survey. Cohort replacement therefore accounts for 7 points of the 13-point rise.

Table 3.  Changes in Support for Same-Sex Marriage

	Change	1985	1997	2003-06
Overall	13	30	38	43
Year of birth:
Before 1940	5	20	23	25
1940s	5	35	40	40
1950s	7	39	39	46
1960s	8	33	42	41
1970s	6	—	45	51
1980s	—	—	—	58
Ideology
Liberal	33	43	66	76
Moderate	13	31	39	44
Conservative	−5	20	21	15
Party identification
Democrat	25	34	47	59
Independent	12	29	41	41
Republican	−3	26	24	23
Level of education
High school or less	10	24	28	34
Some college	18	33	45	41
College graduate	21	43	61	64
Religion
Protestant	4	24	29	28
Catholic	13	25	29	38
Jewish	19	51	61	70
Other religion	20	35	47	55
No religion	25	46	66	71
Race/ethnicity
Non-Hispanic white	14	32	43	46
Latino	11	24	26	35
Black	−1	24	27	23
Asian	30	25	52	55
Gender
Female	14	33	40	47
Male	12	27	35	39

Support has grown much more rapidly among some groups than others, largely reflecting the patterns for individual attitude change. Liberals were already twice as likely as conservatives to favor same-sex marriage in 1985; when their support jumped 33 percentage points while conservatives' dropped 5 points, the gap between them widened to 61 points in 2003-06. The gap between Democrats and Republicans grew from 8 to 36 points, as support rose 25 points among Democrats and dropped 3 points among Republicans. Protestants and African-Americans also hold almost the same positions in 2003-06 as in 1985, leaving them much more opposed to same-sex marriage than members of other religions or race/ethnicities.

A logit model for support for same-sex marriage in 2006 emphasizes the impact of variables only available in that year (Table 4, column 1). Holding the other variables at their means, respondents with close gay friends and others who know gay people are 32 percentage points and 8 points, respectively, more likely to favor same-sex marriage than those who do not know anyone gay. Those who consider religion ‘extremely important’ in their lives are 26 points less likely to favor same-sex marriage than those who consider religion ‘not very important,’ and born-again Christians are 14 points less likely to do so than other Protestants. With those variables in the model, religious affiliation, race/ethnicity, and education do not have significant effects.

Table 4.  Logit Models for Support for Same-Sex Marriage

	1	2	3	4
	2006 Only	2006 Only	1985-2006 Combined	Change in Effect by Survey Period
Notes: Robust z-statistics in parentheses. Coefficients significant at *  10%,** 5%,*** 1%.
Has close gay friend	1.444***
	(6.15)
Has gay acquaintance or relative, but not gay friend	.442*
	(1.93)
Importance of religion in life (1-4)	−.389***
	(4.12)
Born-again Christian	−.674***
	(2.66)
Catholic	−.214	.140	.199*	.199
	(.85)	(.52)	(1.71)	(1.49)
Jewish	.162	.947**	.835***	.191
	(.36)	(2.30)	(4.22)	(.81)
Other non-Christian religion	.063	.526	.538***	.132
	(.21)	(1.88)	(4.11)	(.85)
No religious preference	.246	.982***	1.144***	.327**
	(.83)	(3.71)	(9.39)	(2.44)
Conservatism (1-5)	−.755***	−.816***	−.580***	−.166***
	(5.39)	(5.80)	(12.97)	(4.74)
Republicanism (1-7)	−.368***	−.360***	−.147***	−.129***
	(3.54)	(3.47)	(6.05)	(4.74)
Education (years)	.024	.063**	.064***	.012
	(.80)	(2.24)	(3.59)	(.52)
Male	−.613***	−.682***	−.441***	−.099
	(3.33)	(3.73)	(4.97)	(1.03)
African-American	−.076	−.602	−1.160***	−.563***
	(.18)	(1.57)	(6.07)	(2.76)
Latino	−.181	−.543**	−.703***	−.075
	(.71)	(2.06)	(5.15)	(.44)
Asian-American	−.037	−.316	−.089	.275
	(.12)	(.86)	(.45)	(1.06)
Age (years)	−.029***	−.030***	—	.001
	(4.68)	(4.57)		(.19)
Born in 1920s			−.107
			(.30)
Born in 1930s			.172
			(.50)
Born in 1940s			.758**
			(2.22)
Born in 1950s			.905***
			(2.74)
Born in 1960s			.910***
			(2.70)
Born in 1970s			1.230***
			(3.51)
Born in 1980s			1.685***
			(4.28)
Survey period (1-3)			.128**
			(2.43)
McFadden's pseudo-R^2	.27	.19	.20	.22
Observations	991	991	4,300	4,300

Dropping those variables (column 2) barely alters the coefficients on ideology, party identification, age, and gender. Other coefficients change to become much more like those in 1985-2006 combined (column 3). All the remaining religion coefficients become markedly more positive when religious intensity is not controlled, and when the reference group is all Protestants rather than just those who have not been born again. Minorities become markedly less supportive than comparable non-Hispanic whites, and the impact of education nearly triples and becomes statistically significant. Thus, the differences in support for same-sex marriage across religions, races, and educational levels may well be the result of differences among groups in religious intensity and fundamentalism.

In the model for all years combined, political ideology and age are the strongest predictors of support for same-sex marriage. Holding the other variables at their means, liberals are 27 percentage points more likely than conservatives to support same-sex marriage. Compared to those born in the 1920s and 1930s, support is about 20 points higher for those born in the 1940s, 1950s, and 1960s, and about 30 and 40 points higher for those born in the 1970s and 1980s, respectively. Women are 10 percentage points more likely than men; non-Hispanic whites are 22 points more likely than blacks and 15 points more likely than Latinos; and the nonreligious, Jews, and other non-Christians are 27, 20, and 12 points more likely than Protestants to favor same-sex marriage. Support rises 1.4 points with an additional year of education. Holding all variables at their means, comparable respondents in the most recent surveys are 5 percentage points more likely to favor same-sex marriage than those in 1985. Thus, the net effect of changes in these other variables is to explain the one additional percentage point of the 13-point growth in support for same-sex marriage (in addition to the 7 points already explained by cohort replacement), assuming that the impact of the independent variables has not changed over time.

The impact of several independent variables has strengthened significantly over this period, however. Column 4 shows the coefficients on the interaction terms between survey period and the other independent variables. The most striking changes are the rising impact of party identification and political ideology. Holding the other variables constant, party identification did not matter in 1985, but comparable Democrats and Republicans are deeply divided today. Even holding party constant, the impact of ideology has doubled since 1985. Further, black–white differences were insignificant in 1985 but are highly significant in 2006, and the difference between Protestants and those with no religious affiliation has doubled over this period.

Because the conservatism and partisanship coefficients in the 2006-only models barely change when we drop religious intensity, fundamentalism, and acquaintance with lesbians and gay men, we conclude that coefficients on their interaction terms are “real”: same-sex marriage is an increasingly partisan and ideological issue in California. On the other hand, the 2006 ‘no religious affiliation’ and African-American coefficients are small and statistically insignificant when religious conservatism and intensity are in the model. The other religion coefficients also grow when we drop ‘born-again’ and the ‘importance of religion’ from the model, and all the other religion interaction terms are positive, though insignificant. The real change over time is probably not that those without religious affiliations are increasing their support more rapidly than others. Rather, attitudes toward same-sex marriage probably divide more strongly on religious intensity than they did two decades ago: we expect that those who have been “born again” or consider religion ‘extremely important’ in their lives are at least as strongly opposed to same-sex marriage as they were in 1985, while other Californians are becoming more supportive.

Discussion

Support for same-sex marriage has increased substantially in California. Supporters remain in the minority, but could be become the majority early in the next decade. The change is due partly to individual attitude change and to changing demographics. Age is one of the strongest predictors of opposition to same-sex marriage in California, but not because people are becoming more homophobic or sexually more conservative as they age. In every age group, people are at least four times as likely to say that they have become more accepting of homosexual relations since they turned 18 as they are to say that they have become less accepting. Instead, cohort replacement is the key. Californians born in each decade tend to be more accepting of gay relationships and more willing to grant them legal recognition than those born the decade before—although those born in the 1940s, 1950s, and 1960s are nearly indistinguishable. Cohort replacement by itself explains over half of the increased support for same-sex marriage since 1985.

Although cohort replacement can explain most of the growth, both the individual attitudinal change analysis and the survey-by-survey analysis of support for marriage show that attitude change is concentrated in certain groups. Three patterns stand out. First, gay relationships have become an increasingly ideological and partisan issue in California. Liberals and Democrats are 16 and 18 percentage points more likely than conservatives and Republicans, respectively, to say they have become more accepting of homosexual relations since they were 18. Support for same-sex marriage has jumped 33 percentage points among liberals and 25 points among Democrats, while it has fallen 5 points among conservatives and 3 points among Republicans.

Second, differences based on religious identification and intensity have grown. Members of all religious groups have become more accepting of homosexual relations, but Protestants who have not been born again, Catholics, Jews, and the nonreligious all show net gains of 11 to 17 percentage points greater than born-again Christians. The net gain is 21 percentage points larger for those who said religion was ‘not important’ in their lives than for those who said it was ‘extremely important.’ The gap between Jewish and Protestant respondents on support for same-sex marriage has grown from 27 to 42 percentage points between 1985 and 2003-06, and the gap between those with no religious affiliation and Protestants has jumped from 22 to 43 points. The 2006-only logit model suggests that variables not available for the combined model are the explanation: holding religious affiliation constant, those who consider religion ‘not very important’ in their lives are 26 points more likely to favor same-sex marriage than those who consider religion ‘extremely important,’ and born-again Christians are 14 points less likely to do so than other Protestants, even holding the importance of religion in their lives constant. Born-again Christians and those who consider religion ‘extremely important’ have probably become stronger in their opposition to same-sex marriage while the rest of the state has become more sympathetic.

Third, support for same-sex marriage is not rising among African-Americans as it is among whites, Latinos, and Asian-Americans, and the logit model estimates that the difference in expected support between comparable blacks and whites has tripled between 1985 and 2004. The logit models that control for being born-again, religious intensity, and acquaintance with lesbians and gay men do not show significant black–white differences, suggesting that black–white differences on those three variables explain African-Americans' greater opposition to same-sex marriage. Still, these findings are consistent with growing evidence that support for gay rights is not growing as rapidly among blacks as it is among whites, perhaps because socialization effects are not working in the same way for younger blacks and whites—the differences between older and younger respondents appear weaker within the black community (Allen and Bagozzi 2001; Lewis 2003). Blacks' religious conservatism and the Christian Right's recruitment of black voters specifically on the issue of same-sex marriage may explain the widening black–white gap (Bergeron 2005; Herek 2000; Negy and Eisenman 2005; Smith 1995; Taylor, Mattis, and Chatters 1999).

Implications for the Future of Same-Sex Marriage

Despite strong opposition nationally, several states are moving toward legal recognition of same-sex couples, with equal marriage rights a plausible policy outcome in those states in the next decade (Pinello 2006). Though courts appeared the only avenue for achieving marriage equality a few short years ago, and though legislatures in most states have imposed legal roadblocks to “judicial activism” and state constitutional obstacles to legislative creation of marriage rights, legislatures in a handful of states have created legal rights for same-sex couples, in a context of high salience and simplicity that makes it very difficult for legislators to ignore public opinion. Our research suggests that, in at least one of those states, public opinion has become strikingly more sympathetic to same-sex marriage, providing an opportunity for a lesbian and gay community with more political resources than those in most of the country to pursue marriage equality.

Should we expect to see similar developments throughout the country? On the one hand, cohort replacement accounts for half the rise in support, and older people are dying and children are reaching adulthood in every state. Though the 1977 Field Poll suggests that Californians were already more sympathetic to same-sex marriage than residents of most states two decades ago (see Figure 1), the cohort replacement mechanism could eventually lead to majority support in all states, if succeeding cohorts continue to be more supportive of same-sex marriage. On the other hand, mind-changing in support of same-sex marriage in California is concentrated among liberals, Democrats, and non-Protestants, while born-again and highly religious Christians may be becoming more set in their opposition. California is a very young, “blue,” secular state. It has the fifth lowest median age of the 50 states (Morgan and Morgan 2007). Since 1984, the percentage of the popular vote going to the Democratic presidential candidate has risen from 41 to 54 percent (Leip 2007). Protestant churches, especially more fundamentalist branches, are not as prominent in state politics as in many other states, and it has one of the highest percentage of citizens claiming no religion (Kosmin, Mayer, and Keysar 2001). Even if California's patterns of attitude change hold for the rest of the country, support for same-sex marriage may not rise in states with more politically and religiously conservative citizens, except due to cohort replacement, and cohort replacement may not have the same effect in those states.

We may be witnessing an “issue evolution” on same-sex marriage in California that Lindaman and Haider-Markel (2002) were not able to find at the national level for gay rights more generally. Further research will need to establish whether the divergence on ideological, partisan, and religious lines holds for the rest of the nation when the focus is narrowed from the kinds of questions asked in the GSS (the basis for the Lindaman and Haider-Markel study) to same-sex marriage. California may be a special case, as leaders on both sides of the same-sex marriage battle have successfully mustered their troops. In contrast to many states, California's Democratic leaders have chosen to fight for the relationship rights of their gay and lesbian citizens. With leaders of the Christian Right speaking out loudly against gay marriage everywhere, the debate may be more balanced in California. Democratic voters are getting cues from their leaders that same-sex marriage is something they should care about, and they appear to have been listening as carefully as followers of the Christian Right.

Appendix

Cases

Baehr v. Lewin, 74 Haw. 530, 852 P.2d 44 (1993)

Goodridge v. Department of Public Health, 798 E.2d 941 (Mass. 2003)

Notes

• We would like to thank the Social Science Research and Instructional Council of the California State University System for a Field Institute Faculty Fellowship, which enabled the placement of questions on one of the Field Polls. We would also like to thank Mark DiCamillo of the Field Institute for assistance in the development of the wording of the questions, the inclusion of certain demographic questions on the actual survey instrument, and providing the full results in a timely manner.

• For more information on the subject of same-sex marriage, see the related Website Review: On the Question of Gay Marriage, at the end of this issue of Politics & Policy 36 (1), or visit the online version of this article on our website for the appropriate link.

• 1

  He ranks the top ten in this order: New York, California, Massachusetts, Connecticut, Maryland, Vermont, New Jersey, Virginia, Minnesota, and Hawaii. Of those, five have legal recognition for same-sex partners, and Hawaii has reciprocal beneficiaries.

• 2

  They estimate that 40 percent of Californians would answer yes to the question: ‘Do you think marriages between homosexuals should or should not be recognized by the law as valid, with the same rights as traditional marriages?’ They estimate support to be higher in Massachusetts, Rhode Island, Alaska, New York, New Hampshire, Vermont, and Connecticut.

• 3

  Some experimental and quasiexperimental research examines whether interventions intended to improve heterosexuals' attitudes toward lesbians and gay men succeed, but it has not been terribly successful in discovering effective strategies. See Tucker and Potocky-Tripodi (2006) for a review.

• 4

  We include independents who lean toward one political party as members of that party (following Keith et al. 1992).

• 5

  This was largely because education, gender, and race affected the probability of opinion change rather than the direction of that change. For instance, Latinos reported more positive and negative attitude change than non-Hispanic whites.

• 6

  We measure political ideology on a five-point scale from ‘very liberal’ (1) to ‘very conservative’ (5), and party identification on a seven point scale from ‘strong Democrat’ (1) to ‘strong Republican’ (7).

• 7

  We use the prchange procedure in Stata to generate most probability changes. When interpreting sets of dummy variables (e.g., the religion and political party variables), we set them all to zero and generate the expected probability using prvalue, then change one dummy variable at a time to 1 and calculate the expected probability change. We also use prvalue to compute expected probability differences between college and high school graduates or between 50- and 40-year-olds.

• 8

  This is somewhat surprising, as party identification was not statistically significant in the acceptance of homosexual relations model (column 1) and its impact was only slightly stronger than that of political ideology in the bivariate attitude change analysis (Table 1). Political ideology had the strongest impact in the acceptance model but fell far short of significance in the attitude change model. Multicollinearity is a contributing factor, but ideology is only marginally significant (at the .10 level) and only in the ‘more accepting’ model even when we drop party identification.

About the Authors

Gregory B. Lewis is professor of Public Administration and Urban Studies at Georgia State University and director of the joint Georgia State-Georgia Tech doctoral program in public policy. He has published widely on the career patterns and attitudes of public employees, on public support for lesbian and gay rights, and on morality policy more broadly.

Charles W. Gossett is professor and chair of the Political Science Department at California State Polytechnic University, Pomona. He has published on gay and lesbian issues in state and local politics and in public personnel administration.