The Forum. Volume 9, Issue Article 6. Obama to Blame? African American Surge Voters and the Ban on Same-Sex Marriage in Florida

Transcription

1 The Forum Volume 9, Issue Article 6 Obama to Blame? African American Surge Voters and the Ban on Same-Sex Marriage in Florida Stephanie Slade, American University Daniel A. Smith, University of Florida Recommended Citation: Slade, Stephanie and Smith, Daniel A. (2011) "Obama to Blame? African American Surge Voters and the Ban on Same-Sex Marriage in Florida," The Forum: Vol. 9: Iss. 2, Article 6. DOI: / Available at: Berkeley Electronic Press. All rights reserved.

2 Obama to Blame? African American Surge Voters and the Ban on Same-Sex Marriage in Florida Stephanie Slade and Daniel A. Smith Abstract Did African American voters drawn to the polls by Barack Obama in 2008 cast their ballots for him and then vote to ban gay marriage in Florida, causing Amendment 2 to pass? Using original survey and county-level data, we find the linkage fails to hold. While they tend to be less supportive of marriage equality than whites, black surge voters in Florida were only slightly more likely to support a ban on gay marriage than other likely voters. In addition, we show that although counties that experienced large numbers of black surge voters did exhibit more support for Obama, they were no more supportive of Amendment 2 than other counties. Finally, we demonstrate that black voters were not responsible for carrying Florida s gay marriage ban to passage, as it would have met and surpassed a 60-percent threshold even in their absence. KEYWORDS: African Americans, race, gay marriage, ballot measure, initiative, presidential election Author Notes: Stephanie Slade is a recent graduate of American University, where she became the first person to earn an M.A. in political communications from the School of Public Affairs. Daniel A. Smith is University of Florida Research Foundation Professor and former Director of the Political Campaigning Program. He has published widely on direct democracy in the American states and is the co-author of State and Local Politics: Institutions and Reform. The authors would like to thank the many participants at the 2009 APSA and the 2010 WPSA annual meetings who gave helpful feedback on the paper, as well as those who provided comments when the paper was presented at Penn State University and the University of Florida. Stephanie would like to thank her friend and classmate Zach Moller for his moral and technical support.

3 Slade and Smith: Obama to Blame? 1 Might newly registered voters, mobilized to vote because of a candidate running for office, affect the outcome of a statewide constitutional amendment on the same ballot? More specifically, did Barack Obama s successful candidacy for President in 2008 contribute to the passage of Florida s Amendment 2, with newly registered African Americans turning out to vote for Obama and then casting ballots in favor of the ban on same-sex marriage? If one looks solely at the exit polls, at reporting by the mainstream media, at punditry in the blogosphere, and at self-reflection within the gay community, there is ample reason to believe that such a causal relationship existed in 2008 with respect to the anti-gaymarriage measure in Florida, as well as Proposition 8 in California and Proposition 102 in Arizona. 1 Less than two months before the November 2008 election, a prominent news story appeared in the New York Times suggesting that Obama s popularity among black voters could hurt gay couples (McKinley 2008). According to the account, an unusually high turnout among Obama-inspired minority voters could contribute to the passage of same-sex marriage bans on the ballots in three states, Arizona, California, and Florida. 2 The theory was that newly registered African Americans, as well as those with infrequent vote histories, would surge to the polls in support of Obama, but, due to their traditional social values and religious convictions, would vote for a ban on same-sex marriage, and the theory gained immediate credence. Although as a U.S. Senator, Obama advocated the repeal of the Defense of Marriage Act and opposed a constitutional ban on same-sex marriage, he remained personally opposed to same-sex marriage on religious grounds. Obama did refer to California s statewide ballot measures as unnecessary (Harris 2008), and his acceptance speech at the Democratic convention in Denver did include a statement of support for the gay community (Eleveld and Kennedy 2008), but his campaign assiduously tried to avoid the issue, and the Democratic Party sent only muted cues to oppose the ballot initiatives. Pre-election polling also lent legitimacy to the theory, with a survey from the Public Policy Institute of California finding that a big turnout for the top-of-the-ticket presidential race could have a significant impact on the rest of the ballot (Grew 2008) and an October story in the St. Petersburg Times 1 Blogs with headlines like Obama Voter Turnout Killed Gay Marriage in California (2008) began to surface overnight after the election. The San Jose Mercury News reported just one day after the election that anti-gay-marriage advocates were aided by high minority turnout for Obama (Swift and Webby 2008). Numerous anecdotal explanations for supposed black support for the anti-gay marriage measures surfaced in Florida (see Norman 2008). 2 On November 4, 2008, California s Proposition 8 passed with percent of the vote. In Florida, Amendment 2 garnered percent in a state where a 60-percent super-majority is necessary. Arizona s Proposition 102, a ban similar to California s, passed with percent of the vote. Published by Berkeley Electronic Press, 2011

4 2 The Forum Vol. 9 [2011], No. 2, Article 6 reporting that whether Florida bans gay marriage in its state Constitution could be decided by how much presidential candidate Barack Obama drives turnout among African Americans (Farley 2008). 3 Election-day exit polls only seemed to verify the narrative: Obama s newly-mobilized minority backers, effectively given permission slips by Obama and the Democratic Party to vote for the initiatives, turned out and helped pass the same-sex marriage bans. Drawing on interviews with proponents and opponents of Amendment 2 as well as a mix of public opinion polls and county-level data, we reassess the claim that newly registered African American surge voters were responsible for the narrow passage of Florida s Amendment 2, which codified in the state constitution a law already on the books prohibiting same-sex marriage and civil unions. Specifically, we ask whether black surge voters, who did turn out in droves for Obama in the Sunshine State, really put Amendment 2 over the top. We begin by taking a critical look at the qualification of the initiative for the ballot and the campaign to defeat it. We then situate the surge-voter theory in the broader direct-democracy literature assessing the relationship between ballot measures and candidate races, as well as voter turnout and down-ballot roll-off. Finally, drawing on individual-level survey data and aggregate-level county data from Florida, we put the theory to rigorous empirical tests. The Battle to Ban Gay Marriage in Florida The battle over Amendment 2 began in 2005 when the Florida Coalition to Protect Marriage announced its intent to qualify a constitutional anti-gay-marriage measure, the Florida Marriage Protection Amendment, in time for the 2006 election. Its goal was to collect the 611,009 valid signatures needed to allow voters to decide whether to add a ban on gay marriage to the state constitution. Leading the charge was John Stemberger, an Orlando attorney and President of the Florida Family Policy Council (FFPC), who claimed an amendment was needed to prevent the courts from overturning a 1997 state law prohibiting samesex marriage. The petition proceeded slowly, and the campaign relied on two $150,000 contributions from the Republican Party of Florida to keep its stalled signature drive alive (McMullen 2005). After failing to collect enough signatures to qualify for the 2006 ballot, state officials announced in November of 2007 that Stemberger s organization (renamed Florida4Marriage) had qualified its measure for the ballot in November of The group would eventually raise over $1.5 3 The new story in the St. Petersburg Times was based on a survey conducted for the paper that found 69 percent of black voters supporting Amendment 2, compared to 59 percent of all registered voters in the state and 48 percent of all Democrats. Because of the small sample size, the margin of error for blacks was 9.4 percent (Personal correspondence with author of the story, Robert Farley, October 27, 2008).

5 Slade and Smith: Obama to Blame? 3 million for its campaign, including more than $700,000 of in-kind contributions from FFPC (Kaczor 2008). Parallel campaigns against Amendment 2 soon emerged, one launched by Nadine Smith, the executive director of Equality Florida, the other led by Derek Newton, a Democratic political consultant who ran the ad hoc Florida Red and Blue Coalition. Just as quickly, a rift between the two organizations appeared, stemming from differing political ideologies and a disagreement over strategy namely, whether or not to target minorities. 4 Prior to the election, internal polling for Florida Red and Blue revealed that African Americans overwhelmingly believed homosexuality to be a lifestyle choice and therefore undeserving of protection. The belief was deeply rooted, culturally, Newton said prior to the election, and not addressable by any political campaign. 5 For this reason, his group shied away from the minority community, targeting instead young, unmarried, and more educated voters, who tend to be more tolerant of gays and lesbians (Lewis and Gossett 2008), as well as heterosexual senior citizens. Because the language of Amendment 2 stated not only that marriage is the legal union of one man and one woman but also that no other union that is treated as marriage or the substantial equivalent thereof shall be valid or recognized, opponents worried that its passage would lead to the revocation of domestic partnership rights throughout the state. Cognizant of the successful defeat of a ban on gay marriage in Arizona in 2006, Florida Red and Blue decided to emphasize the amendment s potential implications for all unmarried couples, and in particular, senior citizens. 6 In contrast, Smith contended that Newton s decision not to target African Americans was a mistake. According to Damien Filer, a campaign manager in 2007 for Fairness for All Families (the campaigning arm of Equality Florida), Smith felt strongly that minority voters were not a lost cause. 7 Smith began working in 2006 to make inroads in black communities to oppose the anti-gaymarriage measure, arranging meetings with various progressive and civil rights 4 Smith maintains that the major reason the two campaigns stayed separate was that her organization was perceived as too gay and too progressive (Rodriguez 2008). Members of Florida Red and Blue, which sought to put a bipartisan face on the fight against discrimination, reportedly found Smith to be too militant (Garcia 2008). 5 Author interview with Derek Newton, October 29, Some opponents of Amendment 2 felt that emulating the 2006 Arizona strategy, in which the measure s potential impact on domestic partnerships for senior citizens was put front and center, could be key to defeating Florida s anti-gay-marriage initiative in The Florida Red and Blue campaign distributed flyers noting opposition to the amendment by the League of Women Voters and the Florida Alliance for Retired Americans while pointing out that the initiative could prevent Helene and Wayne, an older heterosexual couple enjoying domestic partnership benefits, from visiting one another in the hospital ( No on 2 group highlights domestic partners 2008). 7 Author interview with Damien Filer, February 12, Published by Berkeley Electronic Press, 2011

6 4 The Forum Vol. 9 [2011], No. 2, Article 6 organizations, and winning endorsements from a number of prominent local African Americans. She felt particularly optimistic that minorities could be persuaded to oppose Amendment 2 as discriminatory if the campaign was tied to Obama s civil rights agenda. To capitalize on Obama s record, Fairness for All Families distributed thousands of handbills and flyers featuring the portrait of the presidential hopeful and urging voters to stand up to discrimination. In Smith s words, people in both white and black precincts gobbled up the Obama cards. Black voters were written off by Florida Red and Blue, Smith lamented, and her own efforts were hindered by a lack of resources. As a grassroots movement, Fairness for All Families had difficulties raising money to wage an effective media campaign, in contrast to Florid Red and Blue, which Smith described as donor-driven by a small cadre of wealthy and bipartisan gay rights supporters. 8 With the passage in November of 2008 of Amendment 2 and Obama s victory in Florida, opponents of the measure were left to wonder whether they had done enough to combat what they saw as a codification of bigotry and a threat to families across the state. Newton said his organization s paid media efforts stood no chance of competing with the concurrent presidential election for voters attention. We were like whispering into a fire drill, he said (Reinhard 2008). Filer, the former campaign liaison for the quarrelling anti-amendment 2 factions, offered another explanation: the gay community was demoralized by the knowledge that same-sex marriage would remain illegal in Florida even if the amendment was defeated. The no campaign lacked a degree of passion and focus at the grassroots, he said, because winning isn t winning. 9 Frustrated by the defeatist consensus within her own camp, Smith claimed that the opponents of Amendment 2 failed effectively to target minority communities, as African American voters could have been moved more. 10 Theorizing about Surge Voter Impact on Ballot Measures Support for the conventional wisdom, that socially conservative black surge voters would make the difference on Amendment 2, was not limited to proponents of the measure. Derek Newton of Florida Red and Blue Coalition said of the theory, Yes. It s true. Still, prior to the election, his campaign s internal polling provided reason to be optimistic, as undecided voters were seen moving to the no side. Newton also pointed out that blacks, historically, have the highest rolloff rate, voting for the top of the ticket candidate but abstaining in down-ballot 8 Author interview with Nadine Smith, February 4, Author interview with Damien Filer, February 12, Author interview with Nadine Smith, February 4,

7 Slade and Smith: Obama to Blame? 5 races or ballot issues. 11 But following the election, Damien Filer, who worked with both opposition campaigns, disagreed with the validity of the theory that a surge of minority voters had led to the passage of Amendment. He said that Florida (unlike Georgia, for example) already tended to experience relatively high turnout among African Americans prior to 2008, due to a concerted effort in previous elections to reach out to minorities. This, he said, created a ceiling on the impact that Obama s candidacy could have on Amendment 2, as there were only so many more blacks in Florida who would not have otherwise turned out to vote. 12 Unfortunately, extant scholarship on direct democracy does little to comment on the soundness of the theory (for a review of the literature, see Smith and Tolbert 2007). Although scholars have found that ballot measures increase turnout (M. Smith 2001; Tolbert and Smith 2005; Lacey 2005), can sometimes increase turnout among peripheral or episodic voters (Donovan, Tolbert, and Smith 2009), and can even affect candidate elections (Nicholson 2005; Smith, DeSantis, and Kassel 2006; Campbell and Monson 2008; Donovan, Tolbert, and Smith 2008), there is also considerable evidence suggesting that the very minority voters who comprised the Obama-inspired surge were likely to have abstained on down-ballot measures like Amendment 2. Using American National Election Study data over several elections to examine patterns of ballot roll-off, that is, voters who turn out to vote for candidates but leave one or more down-ballot races blank, Magleby (1984) finds non-whites to be underrepresented as a proportion of what he refers to as proposition electorates by as much as 25 percentage points. In addition, using a post-election survey, Clubb and Traugott (1972) find that those who voted on ballot propositions in the 1968 presidential election tended to be white and much more highly educated compared to those who voted on candidate races, concluding that statewide referenda constituted an even less effective means to communicate with, and receive communications from, the poor and the poorly educated, blacks, and residents of large cities and rural areas than did other electoral mechanisms. High roll-off rates on initiatives among minority voters thus cast considerable doubt on the surge voter theory. The findings from these studies, however, should not be overstated. In addition to being dated, they may also underestimate the importance of partisan cues, on which some voters are particularly reliant in ballot initiative races (Branton 2003; Smith and Tolbert 2001; 2004). With the presidential nominees of both major parties taking public stances on the proposed gay-marriage bans and the Republican Party of Florida providing early funding of the pro-amendment 2 campaign, cues for party-line voters were not lacking, although it should be noted that the Democratic Party 11 Author interview with Derek Newton, October 29, Author interview with Damien Filer, February 12, Published by Berkeley Electronic Press, 2011

8 6 The Forum Vol. 9 [2011], No. 2, Article 6 offered its supporters little in the way of direction with respect to statewide gay marriage amendments. Despite Obama s opposition to the amendments, for example, he stated publicly that he was not in favor of gay marriage (Harris 2008). Thus, while some party cues existed, party competition on the gay marriage issue was weak at best. Furthermore, Amendment 2 s position as the most salient initiative on the Florida ballot might have mitigated concerns over minority ballot roll-off. Bowler and Donovan (1998: 54) show how some voters are dissuaded from voting on down-ballot issues by the exceptionally high cost of making an informed decision and therefore seek out the more visible contested measures on which to cast their ballots. When information about an issue is easily accessible via wellfinanced campaigns, voters are much less likely to abstain or vote no. In Florida, Amendment 2 enjoyed a distinct advantage in terms of issue salience relative to the five other constitutional amendments on the ballot. Each of the five measures received far less earned and paid media attention than did Amendment 2. Additionally, the number of anti-gay-marriage measures that had appeared on ballots in 2004 and 2006 (Donovan, Tolbert, and Smith 2008; Campbell and Monson 2008), combined with the presence of gay-marriage bans on ballots in California and Arizona in 2008, further increased the salience of the gay-marriage issue in Florida. Thus, Obama-inspired black surge voters may have been less inclined to roll off on Amendment 2 than previous research suggests. Finally, it is possible that election-specific environmental factors may have increased voter awareness of Amendment 2, on which less than half-amillion voters rolled off from the total votes cast for president (a rate of 5.86 percent), perhaps decreasing the likelihood of minority roll-off on the measure. Examining awareness of ballot issues in California, Nicholson (2003) finds that voters are, on average, better than ten percentage points more likely to be aware of ballot measures dealing with morality issues and civil liberties. Campaign spending especially negative advertising corresponds to six percentage points more voter awareness, and media coverage corresponds to an almost seventeen percentage-point increase in awareness. Because Amendment 2 dealt with morality and civil liberties issues, enjoyed considerable attention in the media, and was the subject of two different opposition campaigns that together spent nearly $4 million to influence voter opinion, lower roll-off might reasonably have been be expected, even within demographic groups notorious for high abstention rates. Although non-white voters have been shown to be more likely to roll off on down-ballot initiatives, a wide array of countervailing factors might well have buoyed participation on Amendment 2 among the Obama-inspired surge According to the Office of the Florida Secretary of State, the total number of votes cast for all the presidential candidates in Florida in 2008 was 8,390,744, while the total votes cast either for or against Amendment 2 was 7,898,909. Thus, in reality, nearly half a million voters rolled off on

9 Slade and Smith: Obama to Blame? 7 Scholarship providing insight into the validity of the theory in question is thus inconclusive. Little evidence exists substantiating the notion that a candidate could single-handedly alter the electoral outcome of a ballot measure. Studies pointing to the historically lower turnout rates of non-white voters (Rosenstone and Hansen 1993) and their likelihood to abstain on down-ballot questions (Magleby 1984) cast doubt upon the theory. Other research, however, reveals that Amendment 2 was a candidate for low rates of roll-off, suggesting that the theory s expectations are plausible. African American Surge Voters and Amendment 2: Individual-Level Analysis Dozens of post-election news stories drew upon exit polling to frame the argument that black surge voters were responsible for the passage of the gay marriage ban in Florida. According to exit poll crosstabs as reported by CNN, 71 percent of African American voters in Florida voted for Amendment 2, compared to 60 percent of white voters in the state. 14 Moreover, the exit poll found that even young (18- to 29-year-old) blacks voted for Amendment 2, with greater than 70 percent reporting support for the measure. In comparison, just 49 percent of young whites surveyed in the exit poll reported support for the Florida measure, a number substantially lower than the overall average. Following the election, news outlets proclaimed that newly mobilized black voters carried the anti-gay marriage measure to victory. Yet, when it came to the passage of California s Proposition 8 and the support of minority voters, a post-election study of votes for and against the initiative, funded by the National Gay and Lesbian Task Force Policy Institute, found that support among African Americans had been overstated by the exit polls (Egan and Sherrill 2009). The authors concluded that party identification, ideology, religiosity, and age had a bigger impact than other voter characteristics, including race, on individual propensity to support the measure. In fact, they found that race affected only 5.5 percent of the statewide vote and that black support for Proposition 8 can largely be explained by African Americans higher levels of religiosity. More importantly, the study presented evidence that black support for Proposition 8 was significantly lower than the 70 percent exit poll rate Amendment 2, a rate of 5.86 percent. See, Florida Department of State, Division of Elections, Election Results. Available: 14 As of February 27, 2010, nearly one and a half years after the 2008 presidential election, the exit poll data had yet to be deposited at the Roper Center at the University of Connecticut. For a breakdown of the exit poll marginals, as well as those of the other two surveys, see Appendix A. Exit poll marginals are also available from CNN: /results/polls/#FLI01p1. Published by Berkeley Electronic Press, 2011

10 8 The Forum Vol. 9 [2011], No. 2, Article 6 as reported by CNN and other agencies. The report pointed to the usefulness of combining analyses of precinct- or county-wide numbers with individual-level polling data when attempting to create an unbiased depiction of voter behavior. Survey Data To assess the validity of the surge voter theory and the exit polls in Florida, we draw on two sources of individual data: a 2007 statewide poll of registered voters, as well as the Florida sub-sample of the 2008 Cooperative Congressional Election Survey (CCES). We begin by examining data from an August 2007 pre-election benchmark telephone survey conducted by Lake Research Partners for an opponent of Amendment 2, Fairness for All Families. 15 According to the survey of 1,000 Floridians, which included an over-sample of 200 registered African Americans, blacks did appear to be particularly opposed to gay marriage in Florida. In its memorandum to the campaign, Lake Research Partners reported that likely African American voters supported the amendment by a 6-point margin over likely white voters. Drawing on these data, and controlling for other factors, we explore whether Obama-inspired surge voters were substantially more supportive of Amendment 2 than other likely voters. Since the dependent variable (support or oppose banning gay marriage) is binary, we estimate a logistic regression. A number of individual-level characteristics are known to have the potential to affect attitudes toward gay rights (Rimmerman, Wald, and Wilcox 2000; Haider-Markel and Joslyn 2008; Olson, Cadege, and Harrison 2006). Our logit model includes several control variables: partisanship (with Republicans and Democrats each separately coded 1 and independents coded 0); gender (with women coded 1); age (on an ordinal scale of 1 to 12 where 1 = 18 to 24 years of age, 2 = 25 to 29 years of age, etc.); marital status (with those married coded 1); education (on an ordinal scale of 1 to 6 for the last year of schooling a person completed, where 1 = first through eleventh grades, 2 = high school graduate, 3 = post-high school but non-college education, 4 = some college, 5 = college graduate, 6 = post-graduate education); and religion (with Christians that is, self-reported Catholics, Protestants, and Mormons and weekly churchgoers attendance each separately coded 1). 16 To avoid 15 The RDD phone survey of registered voters was conducted August 14-19, Base sample of 800 likely voters statewide, with oversamples of 100 African Americans (201 African American total including those from the main sample) and 100 Latinos (202 Latinos, total), with a margin of error of +/-3.5 percent. 16 We ran several alternative models, including one also controlling for ballot roll-off (with people who said they always vote on all constitutional amendments coded as 0, and those who said they do not always vote on all amendments coded as 1). In other models (not shown; made available upon request) we included a dummy control variable for Hispanic. The substantive results for all

11 Slade and Smith: Obama to Blame? 9 overstating the statistical significance of the coefficient estimates, we adjust the standard errors by clustering the cases by region (using the state s 10 media markets as a proxy). 17 Given that the August 2007 survey is of registered voters and does not differentiate between surge and non-surge voters per se, we tackle the question of whether African American surge voters were actually more supportive of banning gay marriage by distinguishing between those survey respondents who said they were only probably going to vote in the 2008 general election and those who said they were almost certain to vote. 18 Although this is admittedly an inexact proxy for surge voter status, it does capture the difference between individuals who definitively planned to cast a ballot, regardless of outside factors, and those whose decision about whether or not to turn out was contingent upon developments (such as, perhaps, the Democratic Party s nomination of a black candidate for president). 19 Thus, although the sample of black probable voters may include some non-surge voters, we would not expect the sample of black almost certain voters to include members of the Obama-inspired surge, since these respondents were already committed to turning out long before Obama won his party s nomination. If the surge-voter theory is correct, our expectation is that probable black voters will be more supportive of a gay marriage ban than non-black respondents. In addition to our key independent variable black surge voter we also include in the model dummy variables for black likely voters (i.e., black nonsurge voters), and non-black surge voters, omitting non-black likely voters as our reference category. 20 these models were nearly identical to the base model, and the additional Hispanic variable never reached statistical significance. 17 We also ran the 2007 models (not shown; made available upon request) without clustering by media market, finding no statistical or substantive differences in the results except that the coefficient for marital status was no longer statistically significant. Without the clustering, the coefficients for black surge, black likely, and non-black surge voters remain virtually the same, although the standard errors are lower (p =.803; p =.067; p =.002), respectively. We have chosen to cluster by media markets, as it controls for various campaign effects different regions of the state received different levels of information concerning gay marriage, even prior to the commencement of the media campaign. 18 The initial screening question (Q1) of the survey asks: How likely are you to vote in the November 2008 election for United States President, Congress and other offices are you almost certain to vote, will you probably vote, are the chances about 50-50, are you probably not going to vote, or are you definitely not going to vote? The survey was terminated for those who did not respond almost certain or probably. 19 Using this classification, out of the 1,000 respondents in the 2007 dataset, there are 17 black surge voters, 184 black likely voters, 43 non-black surge voters, and 756 non-black likely voters. 20 We intentionally avoid using interactions (e.g., black * surge voter) in all our logit models, as unlike linear models, the magnitude of the interaction effect in nonlinear models is often Published by Berkeley Electronic Press, 2011

12 10 The Forum Vol. 9 [2011], No. 2, Article 6 Our primary binary dependent variable is generated from responses to the question, Just to be clear, is your vote to ban marriage for gay and lesbian people in the Florida Constitution or not to ban marriage for gay and lesbian people in the Florida Constitution? Respondents who answered they would vote to ban gay marriage are coded 1; all others are coded 0. Although the survey was conducted in 2007, more than a year before the election in question, the gay-marriage issue had already been at the political fore in Florida for two years at the time the poll was taken and Amendment 2 had successfully qualified for the November 2008 ballot, so it is likely that many voters had formulated an opinion on the subject. We expected Republicans to be more likely to support banning gay marriage than Independents or Democrats, and older, married, Christian, and church-going respondents to be more likely as well to support Amendment 2. Because there is considerable literature suggesting minorities are more likely to roll off on ballot propositions than other voters, we include a second model using the August 2007 benchmark poll to test if black surge voters are more likely not to vote on Amendment 2 than other respondents. The model uses identical control variables as our model predicting support for Amendment 2, and we expect black surge voters to be more likely to abstain from voting on the same-sex marriage amendment. We also test the surge voter theory using 2008 CCES pre- and postelection data of more than 1,800 registered and non-registered Floridians. We include a near-identical slate of independent variables as in the model described above (partisanship, marital status, age, education, gender, whether or not the respondent identifies as Christian, and whether or not the respondent attends church weekly). 21 Because the CCES survey does not ask respondents when they registered to vote or what might have motivated them to turn out in the presidential election, we classify surge voters as those who voted in the general election but who were either not registered to vote or who skipped the Democratic and Republican primaries in January. Likely voters are classified as those who voted in both elections. 22 Recall that at the time, the primary election on January 29 in Florida was expected not to count towards selecting the Democratic nominee, as the Democratic National Committee had voted to strip Florida of all its delegates to the 2008 national convention because the state s early primary date violated national party rules. misleading, with signs flipping direction and the marginal effects extremely difficult to interpret (Ai and Norton 2003). 21 Besides the dichotomous surge voter variables, only the ordinal category age in the 2008 CCES data is coded slightly differently from the 2007 Lake Research Partners data (see Appendix A). 22 Using this classification, out of the 1,833 respondents in the 2008 Florida sub-sample that provided answers to these questions, there are 57 black surge voters, 124 black likely voters, 405 non-black surge voters, and 1,247 non-black likely voters.

13 Slade and Smith: Obama to Blame? 11 Our binary dependent variable is support for a federal ban on gay marriage, as the CCES does not specifically ask about Florida s Amendment 2. While not perfect, support for a federal ban on same-sex marriage serves as a proxy for support for Florida s anti-gay marriage amendment, as we feel the average voter is unlikely to distinguish between statewide and federal legislation on this issue. Respondents favoring marriage equality are likely to oppose gay marriage bans at either level, and those opposing same-sex marriage are likely to support any legislation that would ban it. In keeping with the theory, we expect black surge voters in Florida to be more supportive of a nationwide marriage ban than all other likely voters. Evidence of such a relationship would support the notion that the Obama-inspired surge was largely supportive of Amendment 2, and therefore that it may well have contributed to the measure s passage. As with the models using the 2007 Lake Research Partners data, we include two other dummy variables in addition to black surge voter: black likely (i.e., non-surge) voters and non-black surge voters, leaving out non-black likely voters as the reference category. To ensure our classification of black surge voters in fact captures individuals motivated by the Obama candidacy to turn out in the general election (and providing an internal validity check), we estimate a second model with the post-election vote for Obama as the binary dependent variable. Our expectation, in keeping with the first part of the theory, is that black surge voters will be highly supportive of Obama, relative to non-black likely voters. To the extent that Obama mobilized more new voters in the general election than did the Republican Party and John McCain, we also expect non-black surge voters to demonstrate high levels of support for Obama. The dependent variable comes from the 2008 CCES post-election survey in which respondents were asked, For which candidate for President of the United States did you vote? Those who said they voted for Obama were coded 1; those who said they voted for anyone else were coded 0. Survey Findings Model 1 of Table 1 reports the individual-level findings from the 2007 Lake Research Partners logit model predicting support for Amendment As 23 To avoid intermediate variable bias that is, in case other variables besides race and surge status are driving the results, such as Christians and frequent church-goers we ran an alternative stripped-down model (not shown; made available upon request) with the 2007 Lake Research Partners data that included only the three dichotomous surge voter/race variables. In this model, only non-black surge voters are statistically more likely to support Amendment 2 than the average non-black likely voter, with the marginal effect of 11 percentage points higher (p =.039), holding the other two variables at their mean values. The coefficients for the other two variables black surge voters and black likely voters did not achieve standard levels of statistical significance. Published by Berkeley Electronic Press, 2011

14 12 The Forum Vol. 9 [2011], No. 2, Article 6 Table 1.Individual-Level Support for Ban on Gay Marriage and Likely Rolloff Vote in Florida, 2007 Lake Research Partners Data Model 1 Likely Support for Amendment 2 Coef. P-value Marginal Effect Coef. Model 2 Likely Rolloff Voter P-value Marginal Effect Republican (.211) (.122) Democrat (.236) (.124) Married (.106) (.134) Black Surge Voter (.402) (.718) Black Likely Voter (.339) (.292) Non-Black Surge Voter (.389) (.372) Age (ordinal) (.018) (.032) Education (.032) (.065) Female (.226) (.075) Christian (.156) (.110) Church Attendance (.106) (.130) Constant (.339) (.315).041 Pseudo R Log likelihood N Notes: Unstandardized logistic regression coefficients reported, with robust standard errors in parentheses. Standard errors are adjusted by clustering cases by region (10 media markets). Bold coefficients indicate observed statistical reliability at 95 percent confidence intervals. Source: Equality Florida, A Survey of 800 Likely Voters Statewide with Oversamples of 100 African American and 100 Latino Likely Voters, Aug. 2007, telephone survey conducted by Lake Research Partners. expected, Model 1 reveals a positive relationship between Republican identification and support for banning gay marriage and a negative relationship between being a Democrat and support for the constitutional amendment. Education is statistically significant and negative as expected, indicating that better-educated respondents are less likely to support a ban, controlling for other

15 Slade and Smith: Obama to Blame? 13 factors. With respect to religion, Christian respondents, as expected, are more likely to support a ban, and the church attendance covariate (defined as whether a person attends church at least once per week) is also positive and significant. However, the model does not show age or gender to be a good predictor of support for a ban on same-sex marriage. 24 Most important to our study, as Model 1 in Table 1 shows, in the preelection 2007 survey, probable African American voters (our proxy for the Obama-inspired surge) were no more or less likely than all other likely voters to support Amendment 2 at statistically significant levels. This null finding casts doubt upon the conventional wisdom that black surge voters were more supportive of banning gay marriage than whites. In contrast, holding all other independent variables constant, black respondents who self-identified as almost certainly going to vote (i.e., African American likely voters) were likely to support the ban on gay marriage by almost 15 more percentage points than nonblack likely voters, holding all other values at their mean. By way of comparison, Christian respondents were likely to be 17.6 percentage points more supportive of banning gay marriage than those who did not identify as Christians, all else equal. To address the concern about voter roll-off outlined above, we include a second model in Table 1 that tests whether black surge voters were more likely to abstain down-ballot on Amendment 2. As Table 1, Model 2 demonstrates, we find that registered voters who said they were only probably going to vote, regardless of their race, were nearly 38 percentage points more likely to roll off than nonblack likely voters. This finding is consistent with earlier studies that find downballot abstention rates to be higher among non-white and less educated voters (Magleby 1984), and provides evidence to suggest that, whatever their feelings about Amendment 2, members of the Obama-inspired surge were likely to abstain on that issue at a higher rate than likely voters. Less-educated respondents and Christians were also more likely to roll off, although the magnitude of this effect was smaller, at 3 and 4 percentage points, respectively. 24 In an alternative specification of this model (not shown; made available upon request), we control for whether or not (coded 1 or 0) the respondent reported having a close friend and/or family member who was gay. Those who responded in the affirmative were indeed more likely to oppose Amendment 2 by some 19 percentage points, controlling for all other factors. The inclusion of this variable does not notably alter the statistical or substantive findings of the base model we present, and we opt not to include this variable in Model 1 of Table 1 in order to maintain consistency with Model 1 of Table 2 using the 2008 CCES data, as no equivalent question was asked of CCES respondents. Published by Berkeley Electronic Press, 2011

16 14 The Forum Vol. 9 [2011], No. 2, Article 6 Table 2. Individual-Level Support for Federal Ban on Gay Marriage and Support for Obama, 2008 CCES Data Model 1 Likely Support for Federal Ban Model 2 Likely Support for Obama on Gay Marriage Coef. P-value Marginal Marginal Coef. P-value Effect Effect Republican (.164) (.221) Democrat -.668) (.178) (.174) Black Surge Voter (.342) (.508) Black Likely Voter (.312) (.391) Non-Black Surge Voter (.165) (.190) Married (.143) (.163) Age (ordinal) (.097) (.108) Education (.049) (.055) Female (.140) (.160) Christian (.165) (.180) Church Attendance (.156) (.195) Constant (.299) (.351).000 Pseudo R Log likelihood N Notes: Unstandardized logistic regression coefficients reported, with robust standard errors in parentheses. Standard errors are adjusted by sample weights on registered and unregistered voters creating a representative sample of Florida adults. Bold coefficients indicate observed statistical reliability at 95 percent confidence intervals. Source: Stephen Ansolabehere, Cooperative Congressional Election Survey, 2008: Common Content. [Computer File] Release 1: February 2, Cambridge, MA: M.I.T. [producer].

17 Slade and Smith: Obama to Blame? 15 Table 3. Individual-Level Black Support for Amendment 2 (2007 Lake Research Partners data) and Black Support for Federal Ban on Gay Marriage (2008 CCES data) Model 1 Likely Black Support for Amendment 2 (2007 Lake Research Partners Survey) Coef. P-value Marginal Effect.590 (.583) ) (.307) (.186) (.194) (.043) (.093) (.475) (.344) (.297) Model 2 Likely Black Support for Federal Ban on Marriage Coef. P-value Marginal Effect Republican (1.162) Democrat (.447) Surge Voter.183 (.435) Married.670 (.416) Age (ordinal).294 (.300) Education (.146) Female.072 (.447) Christian.479 (.512) Church Attendance (.461) Constant (.354) (.937).904 Pseudo R Log likelihood N Notes: Unstandardized logistic regression coefficients reported, with robust standard errors in parentheses. For Model 1, standard errors are adjusted by clustering cases by region (10 media markets). For Model 2, standard errors are adjusted by sample weights on registered and unregistered voters creating a representative sample of Florida adults. Bold coefficients indicate observed statistical reliability at 95 percent confidence intervals. Sources: Equality Florida, A Survey of 800 Likely Voters Statewide with Oversamples of 100 African American and 100 Latino Likely Voters, Aug. 2007, telephone survey conducted by Lake Research Partners; Stephen Ansolabehere, Cooperative Congressional Election Survey, 2008: Common Content. [Computer File] Release 1: February 2, Cambridge, MA: M.I.T. [producer]. Published by Berkeley Electronic Press, 2011

18 16 The Forum Vol. 9 [2011], No. 2, Article 6 Table 2 reports the results from our two 2008 CCES logit models. 25 As Model 2 reveals, we find (as expected) that both black surge and other surge voters were more likely than non-black likely voters to report having cast a ballot for Obama, holding all other variables constant, while Republicans, Christians, and those who report attending church at least once a week were less likely to support the Democratic Party nominee. In contrast to our earlier findings with the 2007 data, however, Model 1 of Table 2 reveals that both black surge and black likely voters were more supportive of a federal gay-marriage ban. The magnitudes of these results are of particular note, however. Whereas black surge voters were about 21 percentage points more likely to support a gay-marriage ban, the marginal effect for this same group on support for Obama was fully 37 percentage points greater than non-black likely voters. Likewise, black non-surge voters were 20 percentage points more supportive of the federal ban, compared to 33 percentage points more supportive of Obama, relative to non-black likely voters. This indicates that while black voters (both surge and likely) reported more support for a federal ban on gay marriage than did other likely voters, the size of that result was smaller than the size of black voter support for Obama, controlling for other socio-demographic and partisan factors. Stated otherwise, a substantial number of black voters turned out to support Obama (consistent with our expectations) but did not favor a ban on same-sex marriage (contrary to the conventional wisdom). 26 Finally, in Table 3, we report the findings of the same baseline models, but limit the samples to African American respondents only. While the number of respondents in both the 2007 Lake Research Partners and 2008 CCES data is reduced considerably (to 192 and 181, respectively), it is clear from Models 1 and 2 that in neither case does surge voter status result in significantly more or less 25 As with the 2007 data, we ran a model (not shown; made available upon request) with the 2008 CCES data including only the three dichotomous surge/race variables. In this model, only black surge voters are statistically more likely than the reference category of non-black likely voters to support a federal ban on gay marriage, with a marginal effect of 16.2 percentage points higher (p =.065) than the mean non-black likely voter, holding the other two dichotomous variables at their mean values. Coefficients for the other two variables black likely voters and non-black surge voters did not achieve standard levels of statistical significance. 26 We also estimated logistic regression models with both the Lake Research Partners and the CCES datasets that excluded all independent variables other than black surge, black likely, and all other surge voters. In other words, these models (not shown; made available upon request) did not control for additional socio-demographic or political factors such as partisan identification, education, and religiosity. According to the 2007 Lake Research Partners data, black surge voters were still no more likely than non-black likely voters to support Amendment 2; non-black surge voters were 11 percentage points more likely. Conversely, according to the stripped-down 2008 CCES model, black surge voters were 16 percentage points more likely than non-black likely voters to support a federal gay-marriage ban 5 points lower than when the control variables were included.

19 Slade and Smith: Obama to Blame? 17 support among African Americans for a ban on gay marriage (expressed either as support for Amendment 2 or a federal ban on same-sex marriage). These findings suggest that African American surge voters were no different from black likely voters on this issue. Perhaps not surprisingly, only the frequency of church attendance is significant across both datasets African American respondents who attend church at least one time per week were between 12 and 25 percentage points more likely to support a same-sex marriage ban. Our models in Tables 1-3 suggest that support by African Americans for banning gay marriage may have been exaggerated by the press and pundits. Whereas exit polls found that 71 percent of African American respondents in Florida said they voted for Amendment 2 (as we report in Appendix A), our analyses of the 2007 Lake Research Partners and 2008 CCES survey data put that number at 54.2 and 49.2 percent, respectively (also reported in Appendix A). So, why the notable discrepancy? One potential explanation is that survey response bias was stronger in exit polls than in the two phone surveys. Response bias, or the tendency of respondents to give socially acceptable answers to survey questions that fail accurately to reflect their feelings and positions, can make correctly interpreting polling numbers difficult, especially on issues as sensitive as race and homosexuality (Berinsky 2004). If differing social norms or some other factor made non-surge voters inclined to understate their support for Amendment 2 in exit polls, or surge voters inclined to overstate their support, or both, then exit polls may not have accurately reflected these groups relative positions. It also could be that black surge and likely voters included in the exit polls actually rolled off on Amendment 2, but when asked if they voted to ban gay marriage, they reported in the affirmative anyway. Either phenomenon might cause the gap between black and white voters to appear larger than it is, lending undue credence to the belief that high turnout by nonwhite voters aided in the passage of Amendment 2. African American Surge Voters and Amendment 2: Aggregate-Level Analysis Our individual-level analyses suggest that support by black surge voters for a ban on gay marriage differs from support by other likely voters less than was originally reported, if at all. In this section, we delve deeper into the plausibility of the theory that an African American surge was responsible for the passage of the measure by examining aggregate-level support for Obama and Amendment 2, as well as ballot roll-off on Amendment 2. Published by Berkeley Electronic Press, 2011