Finding 1: Which Voters Changed Their Minds


Voter Movement Occurred

Voters other than parents where net movement away from same-sex marriage was substantial:

· Greater Bay Area residents, 31 points;

· Independents, 26 points;

· White Democrats, 24 points;

· Latinos, 10 points

Voters where there was little net movement:

· African-Americans moved 4 points away from us;

· Republicans moved 1 point toward us;

· No on 8 did not carry these groups, but these are not where our support eroded most in the face of the opposition campaign.

Voters moved
: Between mid-September and election day, approximately 5% of voters—at least 687,000— moved to favor the ban on same-sex marriage. The result was a 10-point swing in favor of the ban.In the final six weeks before Election Day, both sides aired $60 million of TV ads and a minimum of 687,000 voters moved to favor the ban on same-sex marriage. More than 500,000 who moved were parents with children under 18 living at home.

Parents accounted for the lion’s share of the movement: Almost three-quarters of the net movement toward the ban was among parents with kids under 18 living at home. Approximately 500,000 of them moved away from us. We lost ground in all voter subgroups likely to contain parents of young kids. The swing away from same-sex marriage occurred among:

·         Mothers, by 26 points

·         Fathers, 20 points

·         Democratic and Independent Mothers under age forty-five, 42 points

·         Democratic and Independent Fathers under age forty-five, 33 points

·         Voters age 30-39, 29 points

·         Voters age 40-49, 21 points

Our newer supporters were more vulnerable: Voters whose support for same-sex marriage was most recent appear to have been the ones most likely to move away in the closing six weeks.


In Prop 8, Voters Moved Away from Same-Sex Marriage

A 10-Point Swing in the Final 6 Weeks

In mid-September, before both sides began to air paid TV advertising, the most reliable and detailed polling available (by Lake Research) showed that Yes and No were neck and neck, with any small apparent lead by one side or the other within the margin of error. Yes on 8 led by 3 points on the issue of same-sex marriage, within the margin of error. At this time, Yes on 8 held commanding leads among the voters over sixty-five and among Republicans. It trailed significantly among the young, especially among young women, and had a huge deficit among white Democrats.

By October 30, when Lake conducted its final tracking poll, Yes on 8 had expanded its lead in favor of the ban on same-sex marriage to 13 points. Exposure to the campaign resulted in 10% more voters opposing same-sex marriage.

Most of the movement away from same-sex marriage happened within groups that were at the start part of No on 8’s base. Yes on 8 made dramatic gains among white Democrats, the most reliably pro-LGBT race-party combination. Yes on 8 made even larger gains in the Greater Bay Area, the strongest area in the state for No on 8. And Yes on 8 made consistent gains among parents and voters of parenting age who were Democrats and Independents.

No on 8’s loss was therefore primarily, though not solely, a function of erosion of support in the final six weeks. During that time, there was a ten-point swing in favor of the ban on same-sex marriage.

The 10-Point Swing = approximately 5% Changed Their Minds

From September 22 on, when TV ads began, approximately 5% of voters changed their minds and switched sides, moving away from support for same-sex marriage to favoring a ban on same-sex marriage. The 5% represents the minimum number who may have changed their minds; it is a close approximation if much or all of the mind-changing was voters starting out supporting same-sex marriage and ending up favoring the ban on it. Alternatively, it is possible that a larger number of voters than 5% changed their minds in a combination of ways, the net result of which was that the ban on same-sex marriage picked up 5% more supporters than our side did. It is probably true that more voters than an exact 5% changed their minds, but we don’t know how many more, because the changes were (or would have been) offsetting, ie an equal number changed in each direction. The data available do not allow calculation of the total number of all voters who changed their minds, only the minimum number that must have to explain the change among the electorate as a whole. The Lake Research data establishes only the floor—the smallest estimate of voters who changed their minds during the campaign.

The Size of the Persuadable Universe of Voters is Unknown, But Could be Far Larger than 5%

The number who changed their minds on same-sex marriage either once or multiple times throughout the final six weeks cannot be determined from the data available, but it could be far larger than 5%. The canvassing experience reported in Recommendation 3, for instance, suggests 13% of those who voted against us on Prop 8 are persuadable. A study by Third Way, a moderate think tank that has studied the recent same-sex marriage ballot measures, leads it to consider the possibility that 46% of the electorate is persuadable. Its report “Moving the Middle on Marriage” is available at http://www.thirdway.org/programs/culture_program/publications/248. Until the LGBT community and our allies conduct more sustained voter persuasion field trials both on same-sex marriage and on the issue of kids, we will not fully know how many voters might be persuadable. Polling alone is not a good enough predictor of voter susceptibility to the kids argument to reveal the degree to which many voters are persuadable in either direction. See Appendices I and L for a more detailed discussion of this topic.

5% May Seem Small, But it is Greater Than the Margin by which the Prop 8 Election was Decided

Some readers may initially think that 5% of the electorate is a very small group of voters. But to put it in perspective, Prop 8 would have had a different outcome if slightly more than 2% of those who voted had switched sides. The estimate of 5% voter movement provided by the Lake data means that the events of the final six weeks decided the outcome. Even if the estimate provided by the Lake data for some reason (unknown to me) overstates the amount of voter movement in the final six weeks and it was merely 4%, that is enough voter movement to have changed the outcome. Even if the estimate provided by the Lake data is wildly exaggerated and voter movement against us in the final six weeks was only 3%, that is enough to have changed the outcome.

This does not mean that Prop 8 could have been won. It is unknown whether an approach different than that taken by No on 8 could have competed more successfully than the approach it did take. It is also possible that the increased salience of the issue of the same-sex marriage as election day came nearer was responsible for some of the voter movement, and that no choice by No on 8 would have availed. As to the latter point, however, some data suggests otherwise: when after a delay No on 8 rebutted the principal Yes on 8 campaign message about kids, some voters backed away from the ban. See Finding 3 for the full discussion of the evidence.

Ultimately, the data show that the final six weeks was decisive. The data are most consistent with the conclusion that the combination of the campaigns waged by both sides in the final six weeks was the proximate cause of the loss for No on 8.


 

Who Were Those 5%?

The following table shows the demographic groups whose support for the ban on same-sex marriage increased the most in the final six weeks of the campaign. The change in opinion is based on the percentage point change between mid-September (when the two sides were in a virtual dead heat) and the last tracking poll on October 30 (by which point Yes on 8 had achieved and sustained its lead).

In the table, the change is labeled "Change in Margin Favoring Eliminate." The table makes a distinction between subgroups with a sample size above and below 100; the smaller the sample size, the more limited the conclusions that we can reliably draw from the data.

Appendix H provides a higher level of detail for many more voter subgroups, both those that moved toward support for the ban and those that moved toward support for same-sex marriage. The text below the table highlights some of the most striking data


 

 

Voter Groups with the Greatest Net Movement Away from Same-Sex Marriage in the Final 6 Weeks

Just to be clear; is your vote to eliminate marriage for gay and lesbian
couples in the state of California or NOT to eliminate marriage for gay or lesbian couples in the state of California?

 

 

30-Oct

22-Sep

October minus September

 

** gray fill indicates a sample size less than 100

Margin Favoring Eliminate

Margin Favoring Eliminate

Change in Margin Favoring Eliminate

Points Gained by Eliminate

Points Gained by Keep

Sample Size >100

Dem/Ind Mothers < 45

0

-42

42

24

-18

Dem/Ind Fathers < 45

29

-4

33

16

-17

Bay Area

-2

-33

31

16

-15

30 - 39

14

-15

29

17

-12

Mother

15

-11

26

13

-13

 

 

 

 

 

 

 

Sample Size < 100

Separated or Divorced

41

-18

59

31

-28

Indep. women

23

-17

40

21

-19

White Indep.

14

-18

32

19

-13

Latino women

18

-8

26

15

-11

35 - 49 Women

6

-16

22

14

-8

 

The bottom line is the center column in dark gray, “Change in Margin Favoring Eliminate.” This column represents how much “Eliminate Marriage’s” lead increased in the final six weeks among the subgroups listed for Lake Polling’s “Be Clear Question.” It is calculated by subtracting the September margin from the October margin, or column 1 minus column 2. The change in margin can also be quantified by looking at the points gained by both sides, or column 4 plus column 5. For a more detailed version of this chart including the raw polling numbers, specific sample size, and movement among other subgroups see Appendix H.


 

Voter Groups with the Greatest Net Movement Away from Same-Sex Marriage in the Final 6 Weeks

No on 8 made progress among voters with postgraduate degrees (they moved 11 points), voters under 30 (6 points) and strong Republicans (4 points). These were the only groups of voters represented by samples of more than 100 voters in the Lake Research polling with movement toward supporting same-sex marriage in the final six weeks greater than the margin of error. For a more detailed version of this chart including the raw polling numbers, specific sample size, and movement among other subgroups see Appendix H.

Just to be clear; is your vote to eliminate marriage for gay and lesbian
couples in the state of California or NOT to eliminate marriage for gay or lesbian couples in the state of California?

 

 

30-Oct

22-Sep

October minus September

 

** gray fill indicates a sample size less than 100

Margin Favoring Eliminate

Margin Favoring Eliminate

Change in Margin Favoring Eliminate

Points Gained by Eliminate

Points Gained by Keep

Sample Size >100

Post-graduate

-16

-5

-11

-3

8

Under 30

-14

-8

-6

0

6

Strong Republican

52

56

-4

-3

1

50 & over Men

15

17

-2

-2

0

Non college men

25

26

-1

1

2

 

 

 

 

 

 

 

Sample Size < 100

Latino 50+

23

38

-15

-9

6

San Diego

10

23

-13

-5

8

Under 35 Men

-3

9

-12

-2

10

(Presidential Vote) Undecided

-4

7

-11

-7

4

Media Market San Diego

11

20

-9

-4

5


Voters Who Moved May be the Most Persuadable

A hypothesis of this report, consistent with my campaign experience, is that it is more relevant to look at the groups that moved away from us in the closing six weeks rather than the voter groups who opposed us in large numbers but who did not move. The very fact that some voters moved away from us means that they were at one point more aligned with us. Their views may therefore be more in flux. While we cannot know which voters are most lastingly persuadable, it is a sure bet that some initial movement is a necessary prerequisite to lasting movement.

Given that the data strongly suggests that Yes on 8’s message worked—it devoted most of its resources to its TV ads; the ads focused largely on fear-mongering about children; and support for Yes on 8 grew after its TV ads gained wide exposure, particularly among the parents it targeted—it’s highly probable that the voters who moved were the most vulnerable to the kids issue. To compete for their support, a future campaign in favor of same sex marriage side will have to respond faster and more effectively to the kids issue if it wants to retain support among the some of the most persuadable voters. It is this thinking that underlies Recommendations 2 and 3 (presented later in this report).

At the other end of the spectrum, it is possible that some voters—perhaps many voters—are not going to change their minds on same-sex marriage. They are for it or against it and that’s that, no matter what campaign ensues. Therefore, identifying voters who are moveable helps us focus.


Key Groups of Voters That Changed Their View on Same-Sex Marriage in the Final Weeks of the Campaign

Parents with children under 18 at home

Approximately 500,000 parent voters with kids under 18 at home moved away from same-sex marriage in the final six weeks.

These parents constituted almost three-quarters of the net movement of all voters away from same-sex marriage during that time.

Although Yes on 8's strongest support was among voters over age fifty, its gains during the campaign came disproportionately from voters under fifty, particularly among parents with kids under 18 at home.

These parents were approximately 30% of all voters in the November 2008 general election. That means that parents cast approximately 4,123,000 of the 13,743,177 total ballots cast on Prop 8.

The Lake polling data shows that a minimum of 5% of all voters moved in the direction of favoring a ban on same-sex marriage in the final six weeks of the campaign. Applying that to the total ballots cast, this report calculates that approximately 687,000 of all voters moved towards favoring the ban.

Similarly, a minimum of 11% of parent voters moved in the direction of favoring a ban on same-sex marriage in the final six weeks of the campaign. Applying that to the number of parent ballot cast, this report calculates that approximately 500,000 of parent voters moved towards favoring the ban. This means that they constituted the large majority of those voters who moved away from us in the final six weeks of the campaign.

For a full description of the data and calculations used to estimate the movement among parent voters, see Appendix H.

Our support deteriorated in every voter subgroup likely to include parents with children under 18 at home.

Same-sex marriage lost ground in every subgroup likely to contain parents with kids under 18. The ban on same-sex marriage gained

·         22 points among parents of children (initially leading by 2 and leading by 24 at the end);

·         29 points among voters 30–39 (initially trailing by 15, leading by 14 at the end);

·         21 points among voters 40–49 (first trailing by 7, then leading by 14);

·         25 points among male voters 35–49 (first trailing by 5, then leading by 20);

·         22 points among female voters 35–49 (first trailing by 16, then leading by 6)

An additional comparison suggests that being a parent is likely to be the single most important independent variable causing voters to move away from No on 8. First, the ban gained 13 points among married women, yet only 9 points among unmarried women; and it gained 7 points among married men while only gaining 2 among unmarried men. Marital status variables have complicated interactions with age (young men, young women, and older women are most likely to be unmarried), but married people are also (when controlling for age) much more likely to be parents.

Chart 1: "Princes" Penetrates and Peels Away Parents
Chart 2: "Princes" Peels Away Parents on the Issue of Same-Sex Marriage

Overall, parents with kids under 18 at home began the campaign evenly divided on same-sex marriage, but ended up against us by a lopsided margin.

Among parents with kids under 18 at home, Yes on 8’s view began ahead in September 46% to 44%, the two-point lead within the margin of error. At the end, opposition to same-sex marriage was 59% to 35%, a 24-point margin.

Yet most of the parents with kids under 18 at home who began the campaign on our side stayed on our side.

Of our initial 44% share of support, we retained 35% at the end. The kids argument cost us 9 points, approximately 20% of the group of parents with which we began. Our share of parents at the end included close to 80% of those with which we began. This allows us to recognize that “Princes” did not work with all parents, and not even with most of the parents who were on our side initially.

Although the sample size is smaller, it appears that parents under age 45 registered as Democrats and Independents were also particularly volatile.

Chart 3: All "Parents" vs Dem & Ind "Parents" Under 45

Although the sample size is small, the data available show that young parents (under forty-five) who are registered as Democrats (D) or Independents (I) started out supporting same-sex marriage. Young mothers were with us 65% to 23%. Young fathers were more closely divided but still on our side 50% to 46%. But support on the issue seriously eroded by the time we got to Election Day. Then, young D & I mothers were tied 47% to 47%; our margin among them had dropped 42 points. Young D & I fathers were against us 62% to 33%, a drop of 33 points.

Chart 3. Correcting for wrong-way voting, support for same-sex marriage deteriorated more among Young Dem & Ind Parents than among Parents as a whole.

Among D & I young dads, support for a ban rose from 33% in May to 62% at the end of Lake polling. This means that one-third of these dads changed their minds on same-sex marriage; or approximately 18% shifted to the opposition. This is a net shift; so if any voters shifted in our direction, the total proportion of those who shifted in one direction or another was greater than that.

The opposition dominated among this group of parents for the last seven days for which Lake Research tracking polling is available. In the final seven days, the opposition went from trailing by 22 to being tied among this group of moms. Similarly, they went from being up 1 point among dads to leading by 29 points. This was a sea change, which in one week about one-eighth of the voters changed their minds (or a quarter changed their minds halfway, eg to/from undecided). This acceleration at the end could well be due to low-information voters disproportionately engaging at the end and being more move-able.

All of the above numbers reflect changes in opinion on the issue of same-sex marriage based on Lake Research’s “be clear” question.

As the above table and the larger one in Appendix H both show, No on 8 had similar problems with all parents and with voters of parenting age. The movement away from same-sex marriage was consistently in the double digits among the wide variety of voter subgroups likely to include large numbers of parents.

The erosion among parents, though not reported until now, should not come as a surprise.

No on 8’s decline among parents was no accident. Yes on 8 ads such as “Princes” were plainly aimed at parents. “Princes” depicts a young daughter seemingly in danger of losing her heterosexuality or her innocence or both. The young girl is oblivious to her peril. She is talking with her mother, who reacts with alarm. Many audiences might empathize with the mother in this situation, but the audiences most likely to see themselves in the ad would be parents, potential parents, and particularly mothers worried about their own children. The same message was reinforced by Yes on 8’s follow-up ads, “Massachusetts” and “Field Trip,” which offered purported facts to bolster “Princes” dramatic narrative. The movement of parents away from No on 8 suggests that the Yes on 8 ads found their target.

Democrats and Independents

Chart 4: Democratic and Independent Voters (Lake Research's Standard Horserace and "Be Clear" Questions)

Chart 5: Democratic & Independent Voters (Lake Research's "Be Clear" Question)

White non-Republicans moved to the anti-LGBT side.

The ban picked up 24 points among white Democrats (the anti–same-sex marriage viewpoint began trailing by 48 and ended trailing by 24).

The ban similarly picked up 32 points (although sample size is small) among white Independents (it started out trailing by 18, but ended carrying this group by 14).

There was essentially no movement among Republicans. On the other hand, there was very little to lose: No on 8 began with only 20% support among Republicans.

It is quite possible that the movement among Democrats mostly reflects movement among the large number of Democrats who are parents and/or of parenting age.

More Whites than Latinos moved to the anti-LGBT side.

Latino and white voters were far from identically situated at the start of the campaign. Whites favored same-sex marriage by 2 points, 46% to 44%; Latinos opposed it by 11points, 49% to 38%. The difference between the two groups was 13 points.

By the end of the campaign, however, the Lake polling showed that whites had flipped and now opposed same-sex marriage by 15 points, 53% to 38%. Latino opposition to same-sex marriage grew and constituted a 21-point margin, 56% to 35%. The gap between the two groups was now only 6 points.

So while Latinos began as less supportive than white voters, and remained less supportive than white voters, the difference between the two groups was cut in half over the course of the campaign from 13 points at the start to 6 points at the end.

Whites and Latinos never became equally supportive, either in their voting on Prop 8 or in their views on same-sex marriage. But by Election Day, Latino voters and white voters were more similar than at the beginning of the campaign.

Women and Men, Latinos and Whites

Women moved away from same-sex marriage. This includes significant movement by Latinas toward the anti-LGBT side.

For the six-week period in which paid campaign advertising aired, the anti–same-sex marriage view gained ground among women. Opposition to same-sex marriage picked up 14 points; it initially trailed by 2 points among women as a whole, and it ended leading by 12 points.

But No on 8 did not lose equal ground among all women. Instead, No on 8’s decline was particularly large among several subgroups.

1.    Among women under fifty, the ban gained 22 points, at first trailing by 22 points and then finally leading by 0;

2.    Among women under thirty-five, the ban gained 20 points, at first trailing by 28 points and then at the end only trailing by 8. Movement among these younger women was particularly striking because by contrast, men under thirty-five moved the opposite direction, toward opposing the ban. See Appendix H for the details not only on young men but also on all male voter subgroups.

Among Latinas, the ban picked up even more support than it did among women as a whole. Latinas switched sides and moved 26 points: they began supporting same-sex marriage by 8 points, and then ended by supporting the ban on same-sex marriage by 18.

By contrast, among Latino men support for the ban on same-sex marriage remained almost unchanged during the final six weeks of the campaign: it went from leading by 25 points to leading by just 24. Lake's sample sizes for Latino subgroups are much smaller than those for women as a whole, so caution must be exercised in interpreting these results.

Again, as with Democrats it is possible that the movement away from No on 8 among women and among Latinas in particular reflect the number of them who are parents or of parenting age.

Based on small samples of both subgroups, it appears likely that younger Latino voters moved toward the anti-LGBT side. Older Latino voters moved toward the pro-LGBT side.

The polling includes only small samples of both subgroups, which means the conclusions must be viewed as tentative. See Appendix H for the specific sample size of each subgroup.

That said, the data indicate a large swing toward support for the ban among younger Latino voters. Initially, among this group the ban trailed by 3 points; but by the end, the ban led by 18.

The data also show a large swing (among an even smaller sample) toward support for same-sex marriage among older Latino voters. They favored a ban at the beginning by 38 points, but supported it by only 23 points at the end.

Chart 6. Latino voters were less supportive of same-sex marriage than whites both at the beginning and the end, but the gap between the two groups is cut in half during the final six weeks because the decline in support among whites exceeds that of Latinos.

Chart 6: Comparing Latino and White Voters

African-Americans and Asian-Pacific-Islanders

Although the No on 8 campaign commissioned a remarkable amount of research, it did not oversample African-Americans or Asian-Pacific-Islanders throughout the campaign. As a result, we know little about changes in opinion among these groups of voters, because the comparative data available includes so few poll participants that the margin of error is larger than any trends in the results.

Among African-Americans, the data we have indicates little movement in the closing weeks of the campaign: initially, the ban led among this group by 9 points, and ended up ahead by 14 points, for a net gain of 5 points. The difficulty is that the initial benchmark poll in May 2008, which included the only oversample of African-American voters, found the ban ahead by 35 points, a very different result. Understanding African-American voters’ views better will have to await additional research. Appendix J, guideline 10 discusses the advantages and costs of obtaining that data consistently in a future campaign.

The situation is similar with Asian and Pacific Islander voters. In the closing weeks of the campaign, the ban went from leading by 7 among API voters to leading by 4. In other words, this was one of the few groups to show movement toward support same-sex marriage by 3 points; this is well within the margin of error, however, so the most accurate way to think about this is that there was no discernible movement in either direction. As with African-Americans, however, the difficulty in accepting these numbers as the totality of the situation is that the benchmark poll in May 2008, which included the only oversample of Asian-Pacific-Islander voters, found the ban ahead by 18 points. Appendix J similarly describes an alternative approach that future campaigns could consider to gain more consistent information.

Geography

The Greater Bay Area moved toward the anti-LGBT side.

Lake’s definition of “Bay Area” includes not only liberal San Francisco, Marin, and Alameda counties, but also Contra Costa, Lake, Napa, San Mateo, Santa Clara, and Sonoma counties. This is broader than the popular conception of the Bay Area, so this report refers to it as the Greater Bay Area. Same-sex marriage lost 31 points in the Greater Bay Area during the campaign. Greater Bay Area voters opposed the ban by 33 points at the beginning, but by only 2 points at the end. They still were with us, but no longer by a big margin.

Education

Less-educated voters moved toward the anti-LGBT side. Based on a small sample of more-educated voters, it appears likely that more-educated voters moved toward the pro-LGBT side.

At the time TV ads began, education was already a significant predictor of voters' attitudes toward same-sex marriage. Less-educated voters favored the ban, while the most educated voters opposed it. But these differences became magnified during the advertising portion of the campaign. The ban gained 17 points among voters with a high school education or less (it initially led by 12, and ended among these voters leading by 29). But the ban lost 11 points among voters who had post-college education (it initially trailed by 5 points, and ended up trailing by 16).

By the end of the campaign, attitudes surrounding same-sex marriage were even more strongly correlated with education than they were at the outset.

Voters whose support of same-sex marriage was more recent

The data suggests that voters whose support for same sex marriage was relatively recent were among the most likely to change their minds in the heat of the Prop 8 campaign. Those last on board were the first off the ship.

It’s striking that the groups of voters moved by Yes on 8 match so closely the groups of voters identified by Lewis and Gossett in their 2008 study “Changing Public Opinion on Same-Sex Marriage: The California Case” (available at http://www3.interscience.wiley.com/cgi-bin/fulltext/119423921/HTMLSTART) as those who in recent years have moved toward support for same-sex marriage. Lewis and Gossett found that liberals, Democrats, whites, and non-Protestants are the groups that have over the last twenty years moved substantially in the direction of supporting same-sex marriage. Lewis and Gossett also found that several other groups (eg Republicans) were moving much more slowly or not at all toward supporting same-sex marriage.

Considering both sources of data together, it seems likely that the groups that have shown the greatest recent movement toward supporting same-sex marriage are the ones most apt to reconsider their position in the heat of a campaign. The newest to support the cause are the easiest for our opponents to peel away.

Similarly, we lost substantial ground with white Democrats. Lake Research anticipated this possibility in a September 19 memo, where it counseled No on 8 to “do better among our base constituencies,” specifically “younger voters (especially women), Democrats, independent women, unmarried voters, and voters in the [Greater] Bay area.” This reinforces the idea that it was within the voter groups where the pro-LGBT side had made gains—where we'd more recently picked up support—that we had a more tenuous hold on voters and suffered the greatest losses.

When it comes to those who moved away from us, it’s not about hate

Taken together, all of the above suggests that those who voted against us who are most persuadable to vote with us next time are not haters. Instead, they are people who have a lot in common with our strongest base of supporters.

Related data from Maine also suggests that it’s not about hate

The situation in Maine may be similar to California. Research performed by Greenberg Quinlan Rosner Research and commissioned by Third Way immediately after the loss on Question 1 showed that there are a group of voters concerned about kids who don’t have a negative attitude toward gay people—and they are mostly Democrats.

Specifically, the polling for Third Way examined the relationship between “LGBT affection” (voters’ overall attitude towards gay and lesbian individuals) and the impact of the “kids are in danger” message. Concern about kids correlated strongly and specifically with two fears: that kids would accept gay couples, and that kids would be raised by gay parents. Concern about kids did not correlate with the specific idea that kids would experiment with being gay. Whether this represents solid data or simply answers that participants were comfortable giving the pollster, they represent, to the best of my knowledge, the first bit of data we have about how moveable voters interpret the “Princes” ad and others like it and why they react to it.

The study found three distinct clusters of voters. The first cluster included a plurality of Republicans: they scored low on “LGBT affection” and they also responded with strong concern when exposed to the “kids are in danger” message. This cluster was not likely to move away from us, because it was never with us. This is consistent with the California numbers that show little Republican movement away from us (or toward us) on Prop 8 in the final six weeks of the campaign.

The second and third clusters of voters each included a large number of Democrats. One group of Democrats scored high on “LGBT affection” and did not respond with concern when exposed to the kids argument. This cluster was also unlikely to move away from us, because it did not find the anti-gay kids ads persuasive. Again, this is consistent with the California experience, where large portions of the Democratic base were the sturdiest in support of same-sex marriage.

The third cluster, however, was a group of Democrats, disproportionately older, who scored high on “LGBT affection,” but who also responded with concern to the kids argument. This cluster included many of those voters most at risk for moving away from us in situation like the Prop 8 and Question 1 campaigns.

The two Greenberg Quinlan Rosner Research scattergrams commissioned by Third Way that reveal the three clusters are below.

Click to view full-size. Reprinted with permission from Third Way.

The overarching point: The Prop 8 campaign shows that a significant number of voters are persuadable on the issue of same-sex marriage late in the campaign.

In 2008 in California, there was a persuadable universe of voters on same-sex marriage. In the final six weeks of the campaign, some voters changed their minds.

This idea that campaigns affect voters would seem self-evident, but it’s worth mentioning in light of the recent report by Prof. Pat Egan of New York University. Some have drawn the conclusion from his report that voters were unaffected by the final six weeks of the Prop 8 campaign and that campaign efforts during that time period do not move voters to change their mind. The San Francisco Chronicle, for example, headlined its story on Prof. Egan’s report this way: “Prop. 8 spending found to have swayed no voters” http://www.sfgate.com/cgi-bin/article.cgi?f=/c/a/2010/06/15/MN051DVGNE.DTL

But to read Prof. Egan’s study that way requires readers to overlook an assumption Prof. Egan made even before he examined the data. His assumption was that all 35 ballot measure campaigns on same-sex marriage were comparable events. Looking at them all together, he concluded that campaigns don't change outcomes.

That assumption overlooks a fundamental difference among the campaigns in his data set; as a result, Prof. Egan was not comparing apples to apples when he looked at the 35 campaigns. Twenty-eight of the campaigns were functionally uncontested: the pro-LGBT side ran very small scale campaigns, with little money, little or no paid mass communications, and infrequent or no direct voter contact. In other words, those pro-LGBT “campaigns” were missing the very elements that give campaigns persuasive potential. See Appendix O for the documentation of the non-competitiveness of these twenty-eight campaigns.

In most states, when the pro-LGBT side mounted modest campaigns, often the anti-gay opposition then followed suit. They realized that once the measure was on the ballot and we didn't fight, they didn't need to fight either. Then they coasted to victory in states where we never had a chance.

Prof. Egan’s conclusion, therefore, does apply to the twenty-eight campaigns that were not seriously contested, but only in a way that is entirely self-evident: when neither side runs much of a campaign, then campaigns don't change outcomes. This report shows, however, that his conclusion is inconsistent with the data on Prop 8 and may not apply to any competitive campaigns. That larger topic awaits further study.

The bottom line is this: When exposed to the Prop 8 campaign, some voters moved. Many more moved away from supporting same-sex marriage than moved toward it. Those who moved were in important number the very voters targeted by the opposition’s message, which appealed to parents to fear for their kids. As the next three findings show, our opposition was smart and effective; we made a few good moves but were less smart and effective; and ultimately, it was our failure to rebut Yes on 8 when it stoked anti-gay prejudice with “Princes” that sealed our defeat.

Methodological Note on Small Voter Subgroups

One example—Democratic and Independent parents under age forty-five—illustrates both the usefulness and the limitations of the available data for small voter subgroups.

Chart 7: All Voters Swing Away from Same-Sex Marriage (Lake Research's "Be Clear" Question)

A great strength of the Lake tracking polling is that it shows how voters as a whole and how subgroups of voters moved during the final six weeks of the Prop 8 campaign.

It is most reliable when sample size is greatest. For example, it is best charting the overall trends among all voters because Lake interviewed so many of them: 400 each day from October 5 through October 30 in daily tracking polling, and then combining three days’ worth of tracking so it could examine data based on a sample size of 1,200. The following chart shows how all voters moved on what we call the Be Clear question, the best single measure of voters’ opinions on the issue of same-sex marriage.

Notice that among all voters, movement on the issue of same-sex marriage is slow and steady in the final weeks of the Prop 8 campaign. Unfortunately, the movement is entirely in the direction of favoring a ban. Later in the report, when we look at the movement among voters on how they planned to vote on Prop 8—the horse race question, which is not the same thing at all as their view on same-sex marriage—there is much more volatility. The comparison and discussion receive extended attention in Finding 2.

In contrast to the Lake data on all voters, its data on smaller groups of voters (eg male voters, female voters, etc.) is both valuable and less fully informative because the smaller the group of voters examined, the larger the potential margin of error.

The Lake data is most limited, most prone to error, when sample size is smallest. In the case of D & I parents under age forty-five with children under eighteen living at home, Lake interviewed only approximately 100-150 of these per poll. The following chart, therefore, is suggestive of the overall trajectory among these voters, but reading too much into mini-trends or one-time daily jumps may simply impute meaning where all that exists is sampling error and anomaly.

Chart 8: All "Parents" vs Dem & Ind "Parents" Under 45

The Lake Research tracking polls suggest that young D & I parents were among those whose movement away from No was partially reversed when “O’Connell” went on the air. The data show that most of these parents started out with us; were affected by “Newsom”; began to succumb to “Princes” after repeated exposure to it and particularly after it had gone unrebutted for almost two weeks; and then some of the defectors returned to our side when we put up “O’Connell” and gave them an answer to the “Princes” argument.

With the small sample size, the following speculation is possible; it is even useful in that it suggests what may have happened; and it suggests areas for further research. The important thing is not to take it as fully proven, since the margin of error is greater than some of the day-to-day changes that the chart depicts.

That said, take a look at the chart. Right at the beginning it suggests that the adverse impact of “Newsom” was larger among D & I young parents than among all voters. By October 2, with “Newsom” on the air against No on 8’s “Thorons” and “Conversation,” the 20-point margin with which we began among young parents had shrunk to 5 points. No on 8 was still ahead, but only by 45% to 40% with 15% undecided. By October 7, with “Princes” on the air along with “Newsom”, refusals spiked to 22%, higher than among voters in general. Both No and Yes lost some voters to undecided at this point; No remained ahead, 42% to 36%.

The chart also suggests that “Princes” was not nearly as effective among D & I young parents in its first week to ten days on the air as it was with voters overall. Over the next week, No on 8 seemed to rebound with the young parents, as most of those who moved to undecided apparently moved back to No. No’s margin most of the week was 13 points. Then the gap closed again, but only briefly, and No regained roughly the same margin for the better part of the next week.

Overall, however, the chart suggests that “Princes” had a big, though delayed, impact on D & I young parents. It could be that the young parents at first resisted “Princes” for more than a week and they backed away from Yes on 8 during this time. But the more time that elapsed without “Princes” receiving a rebuttal, the more that young parents’ support for No on 8 collapsed. By October 19, before No on 8 had aired “O’Connell”, Yes on 8 seemed to take a narrow lead for the first time among D & I young parents of young children, gaining between 1 and 6 points each day. With the sample size so small, the data don’t support a definitive conclusion that Yes actually took a lead. On the other hand, the size of the overall decline in young parents’ support for No on 8 is so great that it is clear that they were not where they had been before TV ads went on the air. As discussed in much more detail in Finding 2, “Princes” was by far the dominant communication reaching these voters from either campaign at this point. This makes it highly likely that the decline among D & I young parents—who were with us only three weeks earlier—was real, and that Yes on 8 had peeled a significant number of them away from No on 8.

 The chart then shows that Yes on 8’s seeming advantage evaporated almost immediately once “O’Connell” went on the air. In the first burst after “O’Connell” aired, D & I young parents almost returned to the level of support of mid-September, before TV ads began.

Further smaller ups and downs that are difficult to attribute to any one cause then transpired. Part of the reason it is difficult to link them to a cause is that in the final stage, so many new ads and earned media emerged almost on top of each other that it is impossible to sort out the influence of one from the other. The first chart, including all voters, suggests that the trend toward Yes on 8 largely continued, but that may not have been true among these younger parents. It’s possible that these parents remained a more volatile voter subgroup until the end. The final Lake tracking polls showed that by October 27, Yes moved into the lead among the younger parents, and it led among them through the time that Lake tracking ended on October 30.

Looking at the voter subgroup data provides food for thought; that’s why we’ve included charts like the one above showing the movement among younger parents for a wide variety of voter subgroups. See Appendix D for the full set of charts. The reason for looking at this one chart in detail, however, is to note not just its value, but also its limitations and the ways in which we have to qualify what we take from the data. In writing this report, I don’t want to offer as proven something that the data only suggest.

Yet it is valuable to glean what we can. A review of the complete timeline shows damage to No on 8. D & I young parents of young children started out as part of the No on 8 base, like D & I voters as a whole. But once Yes on 8 went on the air, young parents moved away, then seesawed back and forth, then partially and temporarily came back when No on 8 rebutted “Princes” with “O’Connell,” only to apparently lose among this group by a narrow margin. It would be easy to make too much of each bobble. On the other hand, it seems likely that “Newsom” and “Princes” put this slice of D & I voters in play.

Yet Democrats and Independents as a whole were overwhelmingly steady and on No on 8’s side throughout the campaign, as the chart above shows. Notice how different its trajectory is from the subgroup of D & I younger parents. If No on 8 had suffered the same decline in support among all Democrats and Independents as it did among the young parents of young kids among them, No would have been crushed. Looking at these two charts together and comparing them to the charts of all voters reinforces our sense that something bad happened among parents, even though our data don’t allow us to understand all of the nuances.

Chart 9: Young Dem & Ind Parents vs. All Dem & Ind Voters

The lesson here for any future campaign is do not assume that voters in the base are fully secured. Consider which subgroups may be particularly vulnerable. We need to prepare our more vulnerable supporters in advance of the closing weeks of the campaign. For example, among younger parents we need to be sure they have the information they need not to succumb to irrational panic about what’s going to happen to their kids. Otherwise we will lose votes we can’t afford to lose.