Changing with the Times

Abstract

Studies have shown that same-sex marriage (SSM) ballot measures affected voter turnout and primed voters in a manner that aided the Republican Party in 2004. However, if attitude strength plays a role in these spillover effects, then recent increases in the intensity of support for SSM on the left may have eroded—or even reversed—the pro-Republican electoral boost of these measures. Using individual- and county-level data, I demonstrate that more recent votes on SSM have mobilized more pro-Obama SSM supporters than pro-Republican social conservatives. These findings are important for understanding how ballot measures may potentially affect candidate elections.

Keywords

same-sex marriage ballot measures voter turnout priming elections

It provided a turnout opportunity for Republicans . . . The belief was, the United States Senate race was not going to be close, and that Republicans needed and social conservatives needed a reason to get to the polls in November.

—Michael Brodkorb, Former MinnesotaGOP Deputy Chair, on the 2012 MinnesotaMarriage Amendment

Liberal democracies are based on a guarantee of basic freedoms to all their citizens. The framers of the Constitution recognized that democratic rule could be tenuous if a majority of citizens could eliminate the basic rights of a minority. This tension motivated the passage of the Bill of Rights (Madison, Hamilton, and Jay 1982). Although the Bill of Rights provided protections at a federal level, state-level processes by which a majority can vote to restrict the rights of a minority remain a potential problem for democracy. The initiative and referenda, as well as the traditional legislative process, can theoretically move society in a less democratic direction if misused. This is especially true if used to amend a state’s constitution and ensure a restriction of rights immune from state judicial review. Some studies have confirmed that minorities, including lesbians and gays, have faced a disadvantage when their rights have come up for a public vote (Haider-Markel, Querze, and Lindaman 2007; D. C. Lewis 2011a). Research also suggests that these votes can cause increased stress on group members, resulting in psychological harm, especially when minorities lose these votes (Russell 2000).

The possibility of restrictions on the civil rights of minorities can be troubling because, at times, it has been good politics for majorities to attack minorities. For example, studies have confirmed that same-sex marriage (SSM) ballot measures in 2004 primed and mobilized voters to the benefit of Republican candidates (Campbell and Monson 2008; Donovan, Tolbert, and Smith 2008). Although the presence of these spillover effects on candidate elections have been contested (Abramowitz 2004; Hillygus and Shields 2008; Jackman 2004), this presumed electoral boost to Republicans has become part of the conventional wisdom regarding these ballot measures and helped to fuel their further proliferation (D. C. Lewis 2011b). To date, thirty states have voted on SSM at least once in the last two decades, and new measures to ban SSM are still in the works (Murray 2013).

Although SSM bans have proved potent in the past, their impact may be less certain at present given an evolving political context. Theoretically, it may take more than just a position in favor of or against an issue on the ballot to motivate voters to either turnout or to use personal positions on the issue as a basis for candidate choice. A voter’s given position on an issue must likely be firm to motivate such an action. It seems unlikely that a voter holding only a weak attitude on an issue will care enough to make a trip to the polls for a public vote on that issue. Likewise, it seems unlikely that a weakly held issue position can serve as an effective prime in a choice between two candidates. Strength of support or opposition should affect the ability of an issue to turnout or prime voters.

Because public votes on SSM have been relatively common over the last decade and attitude strength on this issue has been in flux, SSM allows for a direct test of the importance of attitude strength on these spillover effects. Specifically, attitudes on SSM have become more liberal (Lax and Phillips 2009), and liberals have become much firmer in their support than in 2004 and prior for several reasons. The public has been exposed to SSM more due to increases in social contact with married same-sex couples and increases in depictions of these relationships on film and television. The Democratic Party and liberal mainstream have embraced SSM. This has likely had the effect of increasing the strength of SSM support among liberals and Democrats. Also, the Supreme Court and several lower courts have ruled in favor of same-sex couples seeking legal recognition of their unions. These events have dampened any remaining ambivalence in support for SSM on the left that existed in the years immediately following the initial legalization of SSM in Massachusetts in 2003 (Craig, Martinez, and Kane 2005).

Using data from the 2004, 2008, and 2012 elections, I test whether the priming and mobilization effects of SSM ballot measures have changed in response to shifts in the strength of SSM support and the impact these shifts have had on recent presidential elections. Theoretically, increasing strength in support of SSM on the left should shift these spillover effects from advantaging Republicans to advantaging Democrats. I find that the pro-Republican boost that George W. Bush received from both increases in turnout among religious conservatives in 2004 and from media priming on SSM attitudes in states with public votes on SSM did not materialize for McCain in 2008. Instead I find that SSM ballot measures backfired on Republicans in 2008 by mobilizing more pro-SSM Obama supporters in 2008. To verify this reversal of effects, I then use county-level data to estimate the spillover effects of SSM votes on the presidential election in 2012. I find, in confirmation of the 2008 results, that counties with higher levels of SSM support in states where SSM was voted on had lower rates of drop-off in Obama voters from 2008 to 2012 than states without such measures. One example of such a measure that increased relative Democratic turnout includes the 2012 Minnesota Marriage Amendment mentioned in the quote above and placed on the ballot by Republican state legislators purportedly to aid in the election of Republicans by boosting turnout of social conservatives.

This study has two important findings for the spillover effects of ballot measures on candidate elections in addition to demonstrating that public votes on SSM now aid Democratic candidates: (1) These spillover effects are contingent on attitude strength and (2) because campaigns in support of (or opposition to) a ballot measure are likely to change the public’s attitude strength on a ballot issue over the course of an election, it may be very hard to predict the direction of these spillover effects of in advance especially on issues in which the public initially has given little thought.

Ballot Measures and Candidate Elections

Direct democracy can affect voters in several ways. Itcan increase interest in politics and political knowledge (D. A. Smith and Tolbert 2004; Tolbert, Bowen, and Donovan 2009; Tolbert, Grummel, and Smith 2001; Tolbert, McNeal, and Smith 2003), motivate turnout (Campbell and Monson 2008; Donovan, Tolbert, and Smith 2009; Dyck and Seabrook 2010; Grummel 2008; Lacey 2005; Tolbert and Smith 2005), and alter party identification by framing intergroup conflict (Bowler, Segura, and Nicholson 2006). Central to my purposes, ballot measures can affect turnout in presidential elections (Campbell and Monson 2008) and prime voters to use their positions on specific issues as guidance in candidate choice (Donovan, Tolbert, and Smith 2008; Nicholson 2005; D. A. Smith, DeSantis, and Kissel 2006; D. A. Smith and Tolbert 2010). This combination of effects has the potential to affect election outcomes.

Ballot initiatives involving highly salient morality-based issues like abortion, stem-cell research, and lesbian and gay rights are particularly potent in their ability to encourage turnout (Grummel 2008; Lacey 2005). Regarding SSM specifically, Campbell and Monson (2008) show that SSM ballot initiatives in 2004 motivated evangelical Christians to turnout for GeorgeW. Bush and reduced turnout among secular individuals. These stronger effects are likely because social issues are “easy issues” for the public to comprehend as they are a direct expression of values rather than “hard issues” that require some sort of expertise to understand (Carmines and Stimson 1980).

Scholars are beginning to account for the mechanisms through which ballot initiatives influence the public. D A. Smith and Tolbert (2010) argue that ballot measure campaigns create issue publics or groups of citizens knowledgeable about the ballot issue specifically. They demonstrate that minimum wage ballot initiatives increased the salience of the minimum wage, increased support for changes in the minimum wage, and ultimately primed voters to the aid of Democrats. Donovan, Tolbert, and Smith (2008) show that these effects are concentrated among independent voters, who are disproportionately targeted by political campaigns.

The nature of the election also appears to condition the impact of ballot measures. Peripheral and independent voters are more likely to vote in low-stimulus midterm elections that contain ballot measures but are not more likely to vote in high stimulus presidential elections in which they are already engaged (Donovan, Tolbert, and Smith 2008). These peripheral voters tend to be less sophisticated politically and may be more susceptible to priming effects. They may not care greatly about the candidate elections on the ballot and without the excitement of a high stimulus presidential race to encourage turnout, these voters may require a ballot measure on an issue they care about to motivate a trip to the polls. Taken together, these results imply that both the nature of an election (for instance presidential vs. midterm) and the nature of an issue (for instance, morality vs. policy or easy vs. hard) condition the effects of ballot initiatives on partisan candidate elections.

Despite the general finding that ballot measures have greater effects in midterm elections, research has shown that ballot measures aiming to ban SSM imparted a substantial influence on turnout and priming effects in the 2004 presidential election specifically (Campbell and Monson 2008; Donovan, Tolbert, and Smith 2008). One potential explanation for this contradiction may be that SSM is a quintessentially “easy” issue for the public, so “easy” that it can motivate peripheral voters in both presidential and midterm elections. However, the level of ambivalence and low salience of issues involving lesbian and gay rights (Craig, Martinez, and Kane 2005; Hillygus and Shields 2008) seems to cast doubt on this explanation. An alternative explanation may involve the framing strategies employed by opponents of SSM. By framing SSM as an attack on the “sacred institution” of heterosexual marriage and referring to their position as “pro-family” (Liu and Macedo 2005), opponents of SSM often do not directly refer to lesbians and gays. This may allow individual voters to bypass any ambivalent attitudes they may hold toward lesbians and gays.¹

Rapid changes in SSM attitudes may shed some light on this contradiction in the literature. The recent and sudden growth of support of SSM is what separates it from other political issues. It is on this difference that we should build a theory as the effects of SSM on electoral outcomes and why those effects should change in tandem with public support for SSM.

America’s Shifting Support for SSM

Few issues have displayed as much aggregate change in public support in as short a period of time as SSM. Support has moved from approximately 12 percent in 1988 to more than 50 percent by 2011.² Much of that change has occurred since 2003 (Newport 2011), when Massachusetts became the first state to legalize SSM. This shift is dramatic and striking when compared with the normally glacial tempo of public opinion change in the United States on most issues (Page and Shapiro 1992).

The legalization of SSM in Massachusetts in 2003 was a landmark moment for SSM attitudes in the United States in terms of both attitude change and the issue’s linkage to national politics. The initial legalization of SSM and the subsequent nuptials led to a period of high emotions and issue salience in both 2003 and 2004 (Pinello 2006). Several municipalities across the country also began issuing SSM licenses, in some cases illegally. This drastic change in the status quo was jarring to social conservatives (Pinello 2006). In this context, GeorgeW. Bush began campaigning for “traditional” marriage—marriage limited only to opposite-sex couples.

Although important in 2004, SSM had largely dropped off the broader issue agenda by 2005 and 2006. The Iraq war and the mishandling of hurricane Katrina led to a wave of electoral support for Democrats in 2006. A near global economic collapse caused voters to focus on the economy in 2008. Hence, neither the 2006 nor 2008 elections prominently featured lesbian and gays rights or SSM as a national political issue. Despite this, anti-SSM ballot measures were considered in Tennessee, Virginia, Wisconsin, Arizona, Idaho, South Carolina, South Dakota, and Colorado in 2006 and in California, Florida, and Arizona in 2008.

Although SSM became crowded out by other issues on the national agenda, a number of factors began to push the public in a liberal direction on SSM in the mid to late 2000s. Younger cohorts and those higher in education tend to be more supportive of SSM than others (Brewer 2003; G. B. Lewis and Rogers 1999; Loftus 2001; Wilcox and Wolpert 2000). As those born into newer cohorts and those having more education have come to make up a greater percentage of the electorate, SSM support has increased (Baunach 2012; Loftus 2001). Interpersonal contact with lesbians or gays may also be driving more liberal attitudes (Herek 2002; G. B. Lewis 2007). After Massachusetts, other states such as New York and California³ have legalized SSM. In response, members of the public living in these states may view such legislation as an endorsement of these unions by political leaders and ample research has demonstrated that leaders can influence the positions of their followers (Zaller 1992). Former Governor Arnold Schwarzenegger (R) of California and Governor Andrew Cuomo (D) of New York have endorsed SSM, likely causing their supporters to become more favorable on the issue (Hirshman 2012). In addition, images of same-sex couples in the media are no longer as uncommon as they were in 2003 as television shows like Glee and Modern Family have become popular (Schiappa, Gregg, and Hewes 2006). Finally, many prominent national politicians have also endorsed SSM, including former Presidents Jimmy Carter and Bill Clinton, and former Vice Presidents Al Gore and Dick Cheney.

These numerous factors have certainly had differential effects on the attitude strength of liberals and conservatives regarding SSM. Consequentially, the general level of attitude strength on SSM found among liberals and conservatives may have changed over the course of the 2000s. Elite cues were initially much clearer for conservatives and have likely resulted in less ambivalent opinions early in the evolution of SSM as an issue. On the right, George W. Bush, religious leaders, and other conservative politicians popularized their opposition to SSM in 2003 and 2004. Preexisting opposition to same-sex unions would have strengthened or crystallized in the minds of voters as conservative politicians made their positions clear (Zaller 1992). Outright endorsements of SSM in the intervening years from prominent conservatives (like Vice President Cheney) have been rare. Relatively strong attitudes in opposition to SSM likely provided the basis for turnout and priming effects among conservatives during the 2004 presidential election.

In sharp contrast to conservatives, liberal elites have moved very gradually from tepid support to clear and strong endorsements of SSM over the course of the last decade (Hirshman 2012). The culminations of this trend are recent endorsements of SSM by Vice President Joe Biden and President Barack Obama, which contributed to the relative ease with which the issue was added to the Democratic platform in 2012 (Peters and Shear 2012). Many liberal officeholders opposed SSM in the 1990s and early 2000s, even though their ideological disposition might have pushed them toward a more tolerant position. As liberal elites’ opinions shifted over the course of the 2000s, the nascent support of SSM among those on the left has since increased, and more importantly, strengthened (Baunach 2012). Images of non-threatening same-sex couples in television and film and interpersonal contact with lesbians and gays are also likely reducing ambivalence among liberals and strengthening their attitudes in favor of SSM from 2004 onward. The reduction of this ambivalence, one might argue, was necessary for liberal support of SSM to become sufficiently strong enough to manifest in a priming or turnout effect in the latter half of the decade.

Survey data confirm a shift in intensity of SSM preferences specifically in the context of their importance to presidential elections. Donovan, Tolbert, and Smith (2008) report that Pew’s 2004 mid-August survey showed that 81 percent of those who ranked gay marriage as a “very important” issue in the election were against SSM. The percentage was so high that the authors could safely assume that nearly all those who thought the issue was “very important” were also against SSM. An analysis of Pew’s 2008 mid-October survey, however, shows that this figure had dropped to 66 percent. Although still high, this percentage can no longer be thought of as a proxy for being strongly against SSM. In 2004, 45 percent of those in the Pew survey who were against SSM rated the issue as “very important” in the election, while only 19 percent of those in favor of SSM rated the issue as “very important.” In 2008, 44 percent of respondents opposed to SSM rated the issue as “very important” in the election, while the number in favor of SSM who rated the issue as “very important” had doubled to 35 percent.⁴ This sudden shift should have major effects on how SSM ballot measures affect candidate elections.

SSM Ballot Measures and the Presidential 2008 Election

Changes in the nature of SSM attitudes since 2004 and specifically increases in the strength of support for SSM on the left raise doubts as to whether the pro-Republican effects of anti-SSM ballot measures will be the same as 2004 in subsequent years. Based on changes in the structure of support for SSM, we can draw the following hypothesis:

Hypothesis 1 (H1): Increases in the strength of support for SSM will lead to a decreased ability of SSM ballot initiatives to prime more conservative vote-choice in presidential elections.

There are multiple theoretical reasons as to why this might be the case. There are fewer people strongly against SSM now who can be primed to support Republican candidates. This is especially true of political moderates and Independents who have moved from being strongly against SSM to being more evenly split or supportive in recent years. The remaining pool of voters against SSM is more conservative and Republican than in the past. These individuals would likely be voting Republican regardless of any priming induced by a ballot measures.

The following two hypotheses are related to H1 but involve mobilization rather than priming.

Hypothesis 2 (H2): As time passes, voters opposed to SSM will be less likely to be mobilized based on their SSM attitude as positions against SSM become less strong.

Hypothesis 3 (H3): As time passes, voters in favor of SSM will be more likely to be mobilized based on their SSM attitude as positions in support of SSM become more strong.

A potential rise in ambivalent attitudes due to the factors leading to the mass liberalization on attitudes toward SSM, such as increased contact with lesbians and gays, suggests H2. If voters no longer feel as strongly against SSM as in prior years, they are less likely to flock to polls to oppose it. While social conservatives may not change their attitudes on SSM, the motivation to take the time to vote on a SSM ballot measure would be reduced.

Likewise, H3 is fueled by declining ambivalence and increasing attitude strength among SSM supporters induced by these same reasons. The evidence from the Pew surveys on the electoral importance of SSM supports H3 much more than H2. My expectations are the same. The next hypothesis specifies H2 and H3 for cross-pressured voters.

Hypothesis 4 (H4): H2 and H3 are contingent on candidate support. Voters with candidate choice and SSM attitudes that are congruent will be more likely to turnout than those with incongruent preferences.

This seems reasonable given that cross-pressured voters are less likely to turnout in general, given that they tend to be weak partisans or independents, tend to be less interested in politics, and tend to be more conflicted about voting than those with polarized positions or attitudes.

Taking a broader point of view on these hypotheses in the context of other issues may also clarify the conditions in which ballot measures can affect turnout. One is hard-pressed to find an issue that has demonstrated more aggregate change in public opinion than SSM. The trend has generally involved strengthening support on the left. Most other issues, even morality issues like abortion, do not exhibit this amount of change in attitude strength and ambivalence. Differing levels of ambivalence and attitude strength may themselves be responsible for the differential effects of morality-based ballot measures on turnout.

Testing for Priming Effects in the 2008 Presidential Election

Theoretically, the priming effect of SSM attitudes, which have served to benefit Republicans, should have declined or disappeared since 2004 as the left has become less ambivalent in its support of SSM. To determine if this is the case, I first replicate Donovan, Tolbert, and Smith’s (2008) study, which found that SSM attitudes primed vote-choice in states with SSM ballot measures in 2004. They used a multi-wave panel of Pew polls taken during the 2004 election that gauged candidate choice both before and after the election and asked voters to report the importance of SSM in the election. In 2008, Pew performed a very similar study that asked identical questions. I use this 2008 study to test for priming in the 2008 presidential election.

The first column in Table 1 reproduces the results of SSM issue importance on vote-choice as reported in Donovan, Tolbert, and Smith (2008). I used the same control variables: age, gender, party affiliation, race, education, income, religious denomination, state-level Bush vote in 2000, and respondent assessments of the importance of terrorism, Iraq, and the economy for their presidential vote. The key test for Donovan, Tolbert, and Smith (2008) involved the importance of the SSM issue on a respondent’s vote and the interaction of this with residing in a state with a SSM ballot measure. They hypothesized and found a positive interaction between these two variables on vote-choice. I use a multi-level model clustered by state rather than panel-corrected standard errors as multi-level models are marginally more reliable (Green and Vavreck 2008). As expected, the results are nearly identical to those reported by Donovan, Tolbert, and Smith. The interaction between importance of SSM and voting in a state with a ballot measure (Ban State in 04) is positive and large as previously reported.

Table 1.

Multi-level Logistic Regression: SSM Priming Effects in 2004 versus 2008.

	Voted for Bush (2004)		Voted for McCain (2008)		Voted for McCain (2008)
	Coefficient Estimate	SE	Coefficient Estimate	SE	Coefficient Estimate	SE
Intercept	−4.948***	1.461	−5.548***	1.545	−4.380**	1.625
Ban State in 04 (08)	−0.873*	0.427	−0.692	0.630	−0.863	0.637
Importance of SSM	−0.340	0.406	0.318	0.360
Favor SSM					−1.331***	0.371
Ban State × Importance of SSM	1.897*	0.870	−0.724	0.953
Ban State × Favor SSM					−0.457	0.994
Importance of Terrorism	2.028***	0.422	1.805***	0.361	1.677***	0.372
Importance of Iraq	−1.260**	0.389	−0.114	0.342	−0.070	0.345
Importance of Economy	−0.727*	0.364	−0.747	0.548	−0.537	0.574
Republican	2.614***	0.373	2.288***	0.387	2.311***	0.404
Democrat	−3.162***	0.426	−2.496***	0.396	−2.475***	0.404
Age	0.002	0.010	0.018^†	0.010	0.014	0.010
Male	0.651*	0.297	0.235	0.311	0.044	0.322
White	0.917*	0.446	2.102***	0.457	2.129***	0.446
Education	−0.001	0.109	−0.036	0.137	0.024	0.143
Income	0.080	0.075	0.142*	0.069	0.144*	0.071
Protestant	0.482	0.318	−0.158	0.339	−0.277	0.318
Evangelical	1.355***	0.402	1.006**	0.380	0.788*	0.392
State Bush Vote in 2000 (2004)	5.990**	2.190	2.638	1.949	1.551	2.000
AIC	356.4		337.3		319.3
σ (State)	0.000		0.000		0.000
Deviance	320.4		297.3		283.3
n	554		448		455

SSM = same-sex marriage; AIC = Akaike information criterion.

†

p < .1. *p < .05. **p < .01. ***p < .001 (two-tailed tests).

In the second column of Table 1, I reproduce the same model for the 2008 presidential election using the 2008 Pew data. The key change is that I update the Ban State in 04 dummy variable to measure states with SSM on the ballot in 2008 rather than 2004.

In contrast to 2004, I find no evidence of a priming effect of SSM in 2008. This supports H1. Not only does the coefficient on the interaction term of SSM importance and Ban State fail to reach statistical significance, but it is also in the wrong direction. Race and income were generally more important in 2008. The importance of the Iraq war was less important.

The 2008 Pew study has one clear advantage over the 2004 study for this analysis. As opposed to just asking importance of SSM, the 2008 study also asks whether the respondent favored or opposed SSM.⁵ This allows for a more direct test of priming. In 2004, reported importance of SSM for the present election was highly correlated with an attitude against SSM (Donovan, Tolbert, and Smith 2008), a relationship that faded in 2008 as more individuals have come to favor SSM. In the third column of Table 1, I estimate the same model substituting the respondent’s position on SSM for the respondent’s perceived importance of SSM on their vote. The results are nearly unchanged.⁶

These findings, based on a nearly identical methodology to that which found evidence of priming effects in the 2004 presidential election, show that the priming effects of SSM had disappeared by 2008. However, differences in turnout may have caused these ballot measures to preserve a pro-Republican effect in 2008. In the next section, I test for this mobilization effect.

Testing for Mobilization Effects in the 2008 Presidential Election

The 2008 Cooperative Congressional Election Study (CCES) provides a rich source of data with a large number of respondents. This data set can be used to test for differences in turnout between states with and without SSM ballot measures. In 2008, the study asked respondents specifically how they would vote if their state had a measure banning SSM on the ballot. These questions are used as indicators of a respondent’s position on SSM in 2008. As all respondents were asked this question, it provides a uniform attitude that can be interacted with the presence of a state ballot measure to estimate the effect on turnout. As each voter has three options in the presidential election, “do not vote,” “vote for Obama,” and “vote for McCain,” I use multinomial logistic regression to model vote-choice and turnout simultaneously. “Do not vote” forms the base category. Two equations model turning out and voting for the Democratic candidate or Republican candidate, respectively, rather than not voting.

Because this is an integrated model of turnout and candidate choice, controls for both turnout and vote-choice are included. I use the following variables to control for individual differences in the probability of voting for the Democratic and Republican presidential candidates in 2008: ideology (measured as conservatism), church attendance (religiosity), education, income, political interest, age, and age-squared. I also include dummy variables equaling one if a respondent was a Democrat, Republican, born-again Christian, Protestant, Catholic, female, black, and Latino. I expect income, education, age, church attendance, and political interest to increase turnout generally regardless of vote-choice. Born-again, being a member of a racial minority or a woman, and the square of age (which allows the positive effect of age on turnout to diminish and reverse for the oldest individuals) have generally been shown to lower turnout. I expect Democratic Party identification, liberalism, and being female or a racial minority to predispose individuals to vote for Obama over McCain. I expect Republican identification and greater levels of conservatism to have the opposite effects. I also include state-level fixed effects variables, which control for any state-level variation in the propensity to turn out to vote for either the Republican or Democratic candidates, including those caused by campaigns and campaign intensity at the statewide level. Whenever available, the validated vote from the 2008 CCES is used rather than self-reported vote.

The key variables of interest in this analysis are the dummy variables that represent respondents in states with ballot measures on SSM and the interaction of SSM attitude with the presence of these ballot measures. SSM attitudes are measured as dummy variables taking on the value of one when the respondent takes a position against SSM and zero otherwise. Therefore, a coefficient on Ban State in 08 is the effect of being in a state with a SSM ballot measure on turnout for those favoring SSM. The coefficient on the interaction between voting in a state with a ballot measure and SSM attitude added to the coefficient on the dummy variable for voting in a state with a ballot measure is the effect on turnout for those against SSM.

In Table 2, I estimate a model of vote-choice and turnout in the 2008 presidential election using the 2008 CCES. The coefficients on voting in a state with a ballot measure and the interaction between this variable and holding anti-SSM attitudes are significant in terms of predicting vote-choice for Obama. The estimated coefficients confirm that voters in favor of SSM in states with SSM ballot measures were more likely to turn out and vote for Obama than those in states without such measures. This contrasts with findings from 2004 that showed a net advantage for Republican nominee George W. Bush in 2004. The turnout effect for those who voted for McCain in 2008 and were against SSM is 0.215 (= 0.166 + 0.049). This is roughly half the size of the effect of those in favor of SSM who subsequently voted for Obama (0.383) and also statistically insignificant. The differential effect on turnout of being in support of SSM in a state with a ballot measure on SSM as compared to one without such a vote is similar to that of several years of education or a one-point shift on the five-point ideology scale. This is a moderately sized effect but potentially decisive in a close election because the presence of a ballot measure is one of the few variables in the turnout model that political leaders and activists can directly manipulate using the initiative or referenda process. All other variables in Table 2 behaved as expected with two exceptions. The squared value of age was insignificant statistically and gender (female) was associated with an increased likelihood to vote Republican.

Table 2.

Multinomial Logistic Regression: Turnout and Vote-Choice in the 2008 Presidential Elections (Excluded Category = Did Not Vote).

	Voted for Obama		Voted for McCain
	Coefficient Estimate	SE	Coefficient Estimate	SE
Intercept	−2.328***	0.282	−6.004***	0.287
Against SSM	−0.871***	0.055	0.536***	0.052
Ban State in 08	0.383**	0.159	0.049	0.151
Against SSM × Ban State in 08	−0.257*	0.118	0.166	0.114
Republican	−1.004***	0.073	0.786***	0.049
Democrat	0.874***	0.046	−0.821***	0.063
Ideology	−0.463***	0.026	0.567***	0.027
Religiosity	0.033†	0.020	0.082***	0.019
Born-again	−0.000	0.060	0.156**	0.056
Education	0.117***	0.016	0.046**	0.016
Income	0.061***	0.007	0.087***	0.007
Protestant	0.152**	0.053	0.264***	0.052
Catholic	0.041	0.057	0.394***	0.060
Political interest	0.690***	0.034	0.526***	0.035
Age	0.012^†	0.007	0.023**	0.008
(Age²)/1,000	0.043	0.078	−0.031	0.081
Female	0.029	0.043	0.104*	0.044
Black	1.090***	0.085	−2.043***	0.192
Latino	0.217	0.085	−0.247*	0.097
Open House Seat	−0.007*	0.080	−0.026	0.080
Unopposed House Race	0.109	0.076	0.047	0.081
State-level Fixed-Effects Variables (not shown)
AIC	31,284.58
Deviance	31,004.58
n	23,510

SSM = same-sex marriage; AIC = Akaike information criterion.

†

p < .1. *p < .05. **p < .01. ***p < .001. (two-tailed tests).

Forward: The Effects of SSM Votes in 2012

The results above challenge the conventional wisdom that measures to ban SSM benefit Republican candidates more than Democratic candidates. In this section, I extend these results to the 2012 elections by examining county-level data from states with and without SSM ballot measures. The findings support the individual-level analyses in the preceding two sections. The purpose of this analysis is less about determining whether effects are due to turnout or priming, discussed above, and more about estimating the net spillover effect of these ballot measures in terms of changes in votes cast in states with these measures. Although I set up my models with theoretical turnout effects in mind, modeling shifts in the two party-vote share in a fashion consistent with priming effects rather than changes in votes cast yields identical results.

County-level data were gathered from the four states with SSM ballot measures in 2012: Maryland, Minnesota, Maine, and Washington. All of these were votes to legalize SSM with the exception of Minnesota, which voted on a state constitutional amendment to ban SSM. In sharp contrast to previous years, all four resulted in either the legalization of SSM or the rejection of a ban on SSM. These outcomes provide concrete evidence of change in SSM attitudes. Data were also gathered from all other states that had previously voted on SSM through ballot initiatives or referenda in either the 2004 presidential election or the 2006 midterm elections.⁷ Data on the size of the Hispanic, black, and evangelical populations from each county were also gathered.⁸ Previous evidence from 2004 suggested that evangelical voters may be more likely to be mobilized by these initiatives (Campbell and Monson 2008) and that minority voters were more likely to turnout in 2012 for Obama versus 2008.

Using these data, I modeled the relationship between the percentage in each county that voted to support (or not ban) SSM and the change in either Obama or Romney votes in 2012 from 2008 baselines. My dependent variables here are (1) the ratio of votes cast for Obama in 2012 as compared with 2008 and (2) the ratio of votes cast for Romney in 2012 as compared with McCain in 2008. These variables reflect the number of votes cast in a county for each party in 2012 divided by the number of votes for the same party in 2008. This gives a rough measure of increases or decreases in turnout. These changes in turnout are regressed on the proportion of SSM supporters,⁹ Hispanics, African-Americans, and evangelicals in each county for all the states included. Two dummy variables were created. One represents states with SSM ballot initiatives and one represents swing states. Both were interacted with these four demographic variables to test whether voters in states with ballot measures experienced greater turnout for either Obama or Romney as compared with voters without SSM ballot measures. Dummy variables and interactions with swing states are also included to control for the potential targeting of voters by the presidential campaigns specifically based on SSM attitudes or demographics. Two models were estimated. One involves the effect of SSM support and demographics on increases or decreases in Obama votes from 2008 to 2012, while the other examines increases or decreases in Republican votes. Both of these were estimated using WLS regression. The data were weighted by the square root of each county’s population to correct for heteroskedasticity. The results are presented in Table 3.

Table 3.

Weighted Least Squares (WLS) Regression: 2012 County-Level Change in Vote for President.

	Obama Votes		Romney/McCain Votes
	Coefficient Estimate	SE	Coefficient Estimate	SE
Intercept	0.849***	0.008	1.052***	0.013
% for SSM	−0.003	0.031	−0.173***	0.050
% Hispanic	0.125*	0.055	−0.264**	0.088
% Black	0.339***	0.016	−0.091***	0.025
% Evangelical	0.015	0.022	−0.174***	0.035
Swing State in 2012	0.079***	0.013	−0.006	0.021
Swing × % for SSM vote	0.202***	0.051	0.337***	0.082
Swing × Hispanic	0.154*	0.075	0.354**	0.120
Swing × Black	−0.170***	0.038	−0.146*	0.060
Swing × Evangelical	0.019	0.049	0.368***	0.079
SSM Ballot Measure in 2012	0.113***	0.016	0.039	0.026
SSM BM × % for SSM	0.213**	0.069	0.227*	0.110
SSM BM × Hispanic	0.049	0.088	0.334*	0.139
SSM BM × Black	−0.184**	0.046	−0.049	0.056
SSM BM × Evangelical	−0.049	0.110	0.330†	0.175
Adjusted R²	0.343		0.044
F	55.76		5.93
n	1,497		1,497

Dependent variables are the number of votes cast in 2012 divided by 2008, Weights = sqrt (2010 County Population). SSM = same-sex marriage.

†

p ≤ .1. *p ≤ .05. **p ≤ .01. ***p ≤ .001 (two-tailed tests).

The results are consistent with the individual-level findings from 2008. As shown in column 1 of Table 3, there is no effect of county-level SSM vote on change in Obama vote from 2008 to 2012 in non-swing states without a SSM ban on the ballot. However, in states that held a vote on SSM, there is a statistically discernible and substantive effect as revealed by the interaction of the presence of a ban with county support for SSM. In a county equally divided between SSM supporters and opponents, Obama received about 11.3 percent more votes in states with an SSM vote than in states without a vote. For every 5 percent more or less than 50 percent support for SSM in a county, Obama gained or lost an additional 1 percent of the vote. However, the results are similar not only in states with SSM vote but also in swing states. This suggests that the presidential campaigns may be effectively targeting individuals on either their SSM attitudes or something correlated with these attitudes. Nationally, the larger the Hispanic or black population in a county,the smaller the drop-off in Obama voters. This effect of the size of the Hispanic population on change in Obama vote is larger in swing states, but the effect of the African American population is smaller in both swing states and states with SSM votes. This is because white Democratic turnout was generally higher in these states attenuating the effect of race.

Column 2 in Table 3 shows a different story for changes in votes cast for the Republican nominee. There is a relationship between a change in Republican votes and SSM votes at the county level but not in the expected states. In non-swing states without a 2012 SSM vote, Republican turnout in counties with more opponents of SSM was higher in 2012. However, in states with SSM ballot initiatives, or in states targeted by the presidential campaigns, the interaction coefficients in these models suggest that these pro-Republican effects disappear in the face of higher stimulus elections. The coefficient on the interaction of the presence of a SSM vote with county SSM support is in the wrong direction and cancels the baseline effect found in non-swing states without votes on SSM. The relationship between county support of SSM and Romney turnout is also canceled in swing states by the interaction of the swing state dummy variable and the proportion of county votes favoring SSM. There is no evidence of a statistical relationship between SSM votes and increases in the Republican vote in either swing states or states with SSM votes in 2012 once the marginal effects of the interactions are evaluated (see Kam and Franzese 2007).

Although the model fit for change in Republican votes is generally much weaker and less robust than the Democratic model, additional results emerge. Generally, counties with larger minority populations saw declines in Republican votes nationally. Counties in swing states with larger Hispanic populations, however, did not see a decrease in Republican votes. There is also an unexpected decline in Republican votes in counties with larger evangelical populations. However, this effect was reversed in swing states. Importantly, this decline was likewise reversed in states with SSM votes, but the effect is only marginally significant. This may be due to religious prejudice as Romney was well known to be Mormon. This effect may have been effectively countered by either targeting by Romney campaign or, potentially, a vote on SSM, which may serve to align the interests of evangelical Christians and Mormons. Also, counties with larger Hispanic populations in states with SSM votes also did not see the decline in Republican support seen nationally. On their face, these results appear to signal a potential silver lining for Republicans, suggesting that states with ballot initiatives saw less of a drop in the relationship between the Hispanic population and Republican votes and a stronger relationship between the evangelical population and increases in Republican votes. These two effects may balance or even potentially counter the relationship found in states with ballot measures between an increased relative Obama vote and SSM support.

To test the relative effectiveness of the 2012 ballot measures on affecting the number of major party votes cast, I computed the predicted statewide vote totals with and without the ballot measures in the four states that held them. These results are in Table 4. They are estimated by taking the actual demographics and 2008 vote totals from each county and using these to estimate the number of county-level Obama and Romney votes in 2012 with the SSM ballot dummy variable equaling both 0 and 1. The results were then summated to provide state-level estimates. These findings should be disappointing for Republicans. Because the states that voted on SSM in 2012 have much smaller evangelical and Hispanic populations as compared with individuals who favor SSM (and likely higher levels of those with strong SSM support), the model fits in Table 4 show little net change in Republican votes cast¹⁰ However, the model estimates very substantive increases in Obama votes. These range from an additional net 42,000 votes in Maine to 108,000 votes in Washington. A more modest increase in Obama turnout, 89,000 votes, was estimated for the Republican-backed 2012 Minnesota Marriage Amendment.¹¹

Table 4.

Predicted Vote Totals for States with SSM Votes in 2012 from Models in Table 3.

	Actual votes	Model fit with SSM vote	Model fit withoutSSM vote	Difference in votes	Advantage for Obama
Maine
Obama	399,235	410,000	361,000	48,000
Romney	291,418	312,000	306,000	6,000	+42,000
Maryland
Obama	1,677,844	1,686,000	1,592,000	94,000
Romney	971,869	1,003,000	958,000	45,000	+49,000
Minnesota
Obama	1,546,167	1,546,000	1,380,000	165,000
Romney	1,320,225	1,366,000	1,289,000	77,000	+89,000
Washington
Obama	1,755,396	1,739,000	1,535,000	204,000
Romney	1,290,670	1,322,000	1,225,000	97,000	+108,000

Computed from county demographics, vote totals, and the models in Table 3. SSM = same-sex marriage.

Discussion and Conclusion

With its Hollowsworth versus Perry and United States versus Windsor rulings, the Supreme Court has recently signaled a continued willingness to let SSM play out at a state level for the immediate future.¹² This, coupled with existing statewide constitutional bans on SSM and a newfound liberalism among the public on this issue, ensures that votes on SSM will continue to remain in the public eye. Although SSM measures have benefited Republicans in the past (Campbell and Monson 2008; Donovan, Tolbert, and Smith 2008) and Republican operatives hoping to replicate these effects have placed these measures on the ballot as late as 2012 in Minnesota, I have presented evidence suggesting that these initiatives have disproportionately increased turnout more among SSM supporters since 2008 in response to stronger support for SSM on the left.

As strength in support of SSM has increased rapidly as contrasted with other issues, it has allowed us a glimpse at the differential effects of attitude strength or ambivalence within a single issue rather than across issues. This insight should be taken into account by both researchers and political activists who seek to use ballot measures to stimulate voter turnout. Attitude strength across issues may be largely responsible for higher turnout effects on ballot measures involving morality issues as these issues are easier for citizens and are marked by lower levels of ambivalence. Social or morality issues may not inherently spur greater turnout, but because voters are more confident in their individual positions on these issues, and presumably feel more capable of making an informed choice, these issues may encourage voters to express their opinions more at the polls. Hence, politicians need not target social issues or those involving minority rights if they wish to motivate voters. The key is to find an issue on which public views are firm and unequivocal.

Recent public votes on marijuana legalization can be used as an example to illustrate the issues raised by these findings. Public attitudes on marijuana legalization appear to be in flux, and ballot measures legalizing the possession of marijuana have been discussed as a way to catalyze turnout among younger, more liberal voters. Campaigns in favor of marijuana legalization may create an issue public (D. A. Smith and Tolbert 2010) that will then subsequently turnout and vote for more liberal candidates. However, if attitudes are weakened over the course of a campaign, perhaps due to persuasive or emotional arguments transmitted in the media that raise ambivalence in those initially supporting the legalization, any turnout effect desired by political leaders and activists may evaporate. Similarly, anti-legalization forces may be more effective in strengthening the attitudes of their supporters and motivating them to turnout and (presumable) vote Republican. Because these initiatives must be planned months or years in advance, politically minded proponents who wish to increase turnout must ask themselves the questions “Can public attitudes shift? If so, does this initiative risk potentially helping to elect candidates who oppose my point of view?” These are not easy questions to answer. A vote on a state constitutional ban on SSM, backed by Republican state legislatures, is still moving forward in Indiana (Murray 2013). These legislators should ask themselves these questions as this research suggests that such a public vote would be more likely to turnout voters generally opposed to their interests.

As SSM preferences continue to strengthen among liberals and potentially weaken among conservatives, SSM initiatives will likely continue to advantage liberals in the future with increasing effect. Although the states that held votes on SSM in 2012 generally vote Democratic in presidential elections, pro-SSM activists have recently begun gathering signatures to place new measures legalizing SSM on the ballot in swing states such as Ohio (Staver 2012). Lesbian and gay rights activists, social liberals, and Democrats now appear eager to use the initiative process in an attempt to change the policy status quo on SSM in a more liberal direction. Little fear appears to remain among these individuals in terms of potentially mobilizing social conservatives to the polls and throwing other close elections to socially conservative Republicans. This research helps explain why these fears have changed with the times.

Footnotes

Acknowledgements

The author would like to thank Christian Grose, Bruce Oppenheimer, John Geer, Martin Steinwand, Reuben Kline, the attendees of a poster session at the 2011 meeting of the American Political Science Association, and the anonymous reviewers for their help in improving this article. Marc Hetherington’s suggestions have been particularly valuable. Justine D’Elia, April Strickland, and Jen Sidorova also provided outstanding research assistance.

Declaration of Conflicting Interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abramowitz

Alan

. 2004. “Terrorism, Gay Marriage, and Incumbency: Explaining the Republican Victory in the 2004 Presidential Election.” The Forum 2 (4): Article 3.

Baunach

Dawn M.

2012. “Changing Same-Sex Marriage Attitudes in America from 1988 through 2010.” Public Opinion Quarterly 76 (2): 364–78.

Bowler

Shaun

Segura

Gary

Nicholson

Stephen

. 2006. “Earthquakes and Aftershocks: Race, Direct Democracy, and Partisan Change.” American Journal of Political Science 50:146–59.

Brewer

Paul R.

2003. “The Shifting Foundations of Public Opinion about Gay Rights.” Journal of Politics 6 (4): 1208–20.

Campbell

David

Monson

Quinn

. 2008. “The Religion Card: Gay Marriage and the 2004 Presidential Election.” Public Opinion Quarterly 72:399–419.

Carmines

Edward G.

Stimson

James A.

1980. “Two Faces of Issue Voting.” American Political Science Review 74:78–91.

Craig

Stephen C.

Martinez

Michael D.

Kane

James G.

2005. “Core Values, Value Conflict, and Citizen’s Ambivalence about Gay Rights.” Political Research Quarterly 58 (1): 5–17.

Donovan

Todd

Tolbert

Caroline J.

Smith

Daniel A.

2008. “Priming Presidential Votes by Direct Democracy.” Journal of Politics 70:1217–31.

Donovan

Todd

Tolbert

Caroline J.

Smith

Daniel A.

2009. “Political Engagement, Mobilization, and Direct Democracy.” Public Opinion Quarterly 73 (1): 98–118.

10.

Dyck

Joshua J.

Seabrook

Nicholas R.

2010. “Mobilized by Direct Democracy: Short-Term versus Long-Term Effects and the Geography of Turnout in Ballot Measure Elections.” Social Science Quarterly 91 (1): 188–208.

11.

Green

Donald P.

Vavreck

Lynn

. 2008. “Analysis of Cluster-Randomized Experiments: A Comparison of Alternative Estimation Approaches.” Political Analysis 16 (2): 138–52.

12.

Grummel

John A.

2008. “Morality Politics, Direct Democracy, and Turnout.” State Politics & Policy Quarterly 8 (3): 282–92.

13.

Haider-Markel

Donald P.

Querze

Alana

Lindaman

Kara

. 2007. “Lose, Win, or Draw? A Reexamination of Direct Democracy and Minority Rights.” Political Research Quarterly 60 (2): 304–14.

14.

Herek

Gregory M.

2002. “Gender Gaps in Public Opinion about Lesbians and Gay Men.” Public Opinion Quarterly 66 (1): 40–66.

15.

Hillygus

D. Sunshine

Shields

Todd G.

2008. “Moral Issues and Voter Decision Making in the 2004 Presidential Election.” PS: Political Science & Politics 38 (2): 201–209.

16.

Hirshman

Linda

. 2012. Victory: The Triumphant Gay Revolution. New York: Harper Collins.

17.

Jackman

Simon

. 2004. “Same-Sex Marriage Ballot Initiatives and Conservative Mobilization in the 2004 Election.” http://jackman.stanford.edu/papers/risspresentation.pdf.

18.

Kam

Cindy D.

Franzese

Robert J.

Jr.

2007. Modeling and Interpretation of Interactive Hypotheses in Regression Analysis. Ann Arbor: University of Michigan Press.

19.

Lacey

Robert

. 2005. “The Electoral Allure of Direct Democracy: The Effect of Initiative Salience on Voting, 1990–1996.” State Politics & Policy Quarterly 5:168–81.

20.

Lax

Jeffrey R.

Phillips

Justin H.

2009. “Gay Rights in the States: Public Opinion and Policy Responsiveness.” American Political Science Review 103:367–86.

21.

Lewis

Daniel C.

2011a. “Bypassing the Representational Filter? Minority Rights Policies under Direct Democracy Institutions in the U.S. States.” State Politics & Policy Quarterly 11:198–222.

22.

Lewis

Daniel C.

2011b. “Direct Democracy and Minority Rights: Same-Sex Marriage Bans in the American States.” Social Science Quarterly 92:375–94.

23.

Lewis

Gregory B.

2007. “Personal Relationships and Support for Gay Rights.” Andrew Young School of Policy Studies Research Paper Series No. 07-10. http://ssrn.com/abstract=975975.

24.

Lewis

Gregory B.

Rogers

Marc A

. 1999. “Does the Public Support Equal Employment Rights for Gays and Lesbians?” In Gays and Lesbians in the Democratic Process, edited by Riggle

Ellen D. B.

Tadlock

Barry L.

, 118–45. New York: Columbia University Press.

25.

Liu

Frederick

Macedo

Stephen

. 2005. “The Federal Marriage Amendment and the Strange Evolution of the Conservative Case against Gay Marriage.” PS: Political Science & Politics 38:211–15.

26.

Loftus

Jeni

. 2001. “America’s Liberalization in Attitudes toward Homosexuality, 1973 to 1998.” American Sociological Review 66 (5): 762–82.

27.

Madison

James

Hamilton

Alexander

Jay

John

. 1982. The Federalist Papers. New York: Bantam Books (Bantam Classic edition).

28.

Murray

Jon

. 2013. “Freedom Indiana coalition kicks off fight against gay marriage ban amendment.” Indianapolis Star. Aug. 26th. http://www.indystar.com/story/news/politics/2013/08/21/freedom-indiana-coalition-kicks-off-fight-against-gay-marriage-ban-amendment/2681503/

29.

Newport

Frank

. 2011. “For First Time, Majority of Americans Favor Legal Gay Marriage.” Gallup, May 20. http://www.gallup.com/poll/147662/first-time-majority-americans-favor-legal-gay-marriage.aspx.

30.

Nicholson

Stephen P.

2005. Voting the Agenda: Candidates, Elections, and Ballot Propositions. Princeton: Princeton University Press.

31.

Page

Benjamin I.

Shapiro

Robert Y.

1992. The Rational Public. Chicago: University of Chicago Press.

32.

Peters

Jeremy W.

Shear

Michael D.

2012. “Democrats Draft Gay Marriage Platform.” The New York Times, July 30, A14.

33.

Pinello

Daniel R.

2006. America’s Struggle for Same-Sex Marriage. New York: Cambridge University Press.

34.

Russell

Glenda M.

2000. Voted Out: The Psychological Consequences of Anti-Gay Politics. New York: New York University Press.

35.

Schiappa

Edward

Gregg

Peter B.

Hewes

Dean E.

2006. “Can One TV Show Make a Difference? Will and Grace and the Parasocial Contact Hypothesis.” Journal of Homosexuality 51 (4): 15–37.

36.

Smith

Daniel A.

DeSantis

Matthew

Kissel

Jason

. 2006. “Same-Sex Marriage Ballet Measures and the 2004 Presidential Election.” State and Local Government Review 38 (2): 78–91.

37.

Smith

Daniel A.

Tolbert

Caroline

. 2004. Educated by Initiative: The Effects of Direct Democracy on Citizens and Political Organizations in the American States. Ann Arbor: University of Michigan Press.

38.

Smith

Daniel A.

Tolbert

Caroline

. 2010. “Direct Democracy, Public Opinion, and Candidate Choice.” Public Opinion Quarterly 74 (1): 85–109.

39.

Smith

Tom W.

Marsden

Peter

Hout

Michael

Kim

Jibum

. 2011. General Social Surveys, 1972–2010. Cumulative Codebook. Machine-Readable Data File. NORC ed. Chicago: National Opinion Research Center; Storrs, CT: The Roper Center for Public Opinion Research, University of Connecticut.

40.

Staver

Anna

. 2012. “Same-Sex Marriage Amendment in Ohio Gets Green Light.” Huffpost Politics, April 3. http://www.huffingtonpost.com/2012/04/03/same-sex-marriage-amendment-ohio_n_1400714.html.

41.

Tolbert

Caroline J.

Bowen

Daniel

Donovan

Todd

. 2009. “Initiative Campaigns: Direct Democracy and Voter Mobilization.” American Politics Research 37 (1): 155–92.

42.

Tolbert

Caroline J.

Grummel

John

Smith

Daniel

. 2001. “The Effects of Ballot Initiatives on Voter Turnout in the United States.” American Politics Research 29:625–48.

43.

Tolbert

Caroline J.

McNeal

Ramona S.

Smith

Daniel A.

2003. “Enhancing Civic Engagement: The Effect of Direct Democracy on Political Participation and Knowledge.” State Politics & Policy Quarterly 3:23–41.

44.

Tolbert

Caroline J.

Smith

Daniel

. 2005. “The Educative Effects of Ballot Initiatives on Voter Turnout.” American Politics Research 33:283–309.

45.

Wilcox

Clyde

Wolpert

Robin

. 2000. “Gay Rights in the Public Sphere: Public Opinion on Gay and Lesbian Equality.” In The Politics of Gay Rights, edited by Rimmerman

CraigA.

Wald

Kenneth D.

Wilcox pgs

Clyde

. 409-432. Chicago: University of Chicago Press.

46.

Zaller

John

. 1992. The Nature and Origins of Mass Opinion. New York: Cambridge University Press.