This Appendix uses graphs and charts to illustrate shifts in public opinion during the Prop 8 campaign. In a few cases, it also uses charts to demonstrate changes that occur in the larger political environment, such as in earned media coverage, in the final six weeks of the campaign.
At a glance, each uses the data to illuminate a trend. All together, the charts communicate the gestalt—the big picture—and reveal the relationship between key campaign events and the way voters responded to them.
The charts remedy a potential deficiency of the rest of the report, its length. In this section, the same important ideas are shared, shorn of some detail and text. The result: a clear picture of the largest events that affected the fortunes of No on 8.
The tables and graphs in this appendix are similar to those throughout the report, but here they are larger and easier to read.
Most charts and graphs in this Appendix use data collected by the two polling firms hired by the No on 8 campaign: Lake Research and David Binder Research. Lake Research polled for a longer period of time, from May through October 30, and it polled more frequently. David Binder Research first polled for No on 8 September 2–4, and then returned to do daily tracking only toward the very end of the campaign.
In addition, some charts draw upon polling done by respected statewide and national polling firms that polled on Prop 8, including the Public Policy Institute of California (PPIC) and the Field Poll, though their surveys were at much more sporadic intervals than either of the two firms commissioned by the campaign.
Finally, a few charts present data that does not come from polling at all. For example, data gleaned from targeted searches of the Google News Archive measure trends and help reveal trends in earned media coverage of the campaign.
The tables and graphs below are grouped by poll question or topic. For example, one of Lake Research’s questions, the Standard Horse Race, is presented for all voters, and then separately for several subgroups of voters.
Where possible, both tables and graphs are presented for all data sets. A few paragraphs of text before the graphs (and text boxes superimposed on the graphs) describe the source of the data, its relevance, conclusions this report draws from the data, and any deficiencies in the data. In addition, a summarized chronology of events is below each chart to provide context.
Throughout most of this appendix, tables and graphs are grouped together by topic. Here, however, the charts which tell the most impactful stories in the simplest ways are presented together. This way, you don’t have to go digging through charts which are often more complex and not as easily decipherable to find the most compelling pieces of data.
Lake Research’s Standard Horse Race Question
Lake Research was the main polling firm used by the No on 8 campaign. Lake conducted benchmark polls—longer polls that asked fifteen to thirty minutes of questions—in May, July, August, and September. Each took place over a week and surveyed around 800 voters. On September 22, Lake began conducting much shorter tracking polls. The results of the tracking polls, taken daily, were released in three-day rolling averages, a standard practice: numbers from three concurrent days were combined, both to provide a larger and more reliable sample size (around 1,000) and to average out potentially misleading day-to-day fluctuations and more reliably show trends.
The No on 8 campaign used Lake’s polls to test both sides’ expected campaign messages, to gain some idea of what voters were hearing about the campaign, and to ascertain in as much depth as possible what voters thought about same-sex marriage and the ballot measure. In addition, Lake gathered extensive information about survey participants, including gender, race, party affiliation, and other demographics. Lake also intentionally oversampled Latino voters. Oversampling compensates for the fact that Latino voters do not comprise a large enough proportion of likely voters to be broken down further and still have a reliable sample size to detect trends among subgroups of Latino voters. By oversampling, Lake intentionally polled more Latino voters than would be present in a random sample, obtaining a large enough Latino sample to permit more sophisticated and detailed analysis of Latino voters.
The Standard Horse Race question referred to here and throughout the Report is Lake’s standard tracking question using the Prop 8 ballot language. The question reads:
Proposition 8 on the November ballot is an Initiative Constitutional Amendment titled, “Eliminates Rights of Same-Sex Couples to Marry.” It changes the California Constitution to eliminate the right of same-sex couples to marry. It provides that only marriage between a man and a woman is valid or recognized in California. The fiscal impact over the next few years is potential revenue loss, mainly sales taxes, totaling in the several tens of millions of dollars, to state and local governments. In the long run, there is likely to be little fiscal impact on state and local governments. If the election were today, would you vote yes or no on Proposition 8?
[IF YES/NO] Do you feel that way strongly or not-so-strongly?
[IF UNDECIDED] Well, if you had to decide today, would you lean toward voting yes or would you lean toward voting no?
This Appendix first examines the Standard Horse Race response from all voters surveyed. Responses to the Standard Horse Race question were also broken down by gender and race (presented later). All results are located on a time line to show the results in relationship to campaign events. As mentioned elsewhere, the main focus of both sides was paid media, and so paid media is the focus of the time line.
There are two factors that keep the Standard Horse Race question from being a completely reliable source of data about voters’ opinions on the issue of same-sex marriage. The first and most important is voter confusion: voters’ intentions as to whether they wished to allow or ban same-sex marriage often differed from their reported yes or no on Prop 8 reponse to the Standard Horse Race question. The Standard Horse Race question therefore combines and reflects three separate trends: (1) how voters plan to cast their vote on Prop 8; (2) fluctuating levels of voter confusion on Prop 8; and (3) whether voters support or oppose same-sex marriage.
Another limitation of the Standard Horse Race question is the grouping of undecided, refused, and don’t know responses in one category. This is discussed after the following graphs in the section entitled “Lake Research’s Standard Horse Race for Voters Yet to Vote.”
Lake Research’s Standard Horse Race for Voters Yet to Vote
Lake Research broke down the responses to its Standard Horse Race question in different ways, depending upon the kind of voter responding. Prior to October 6, all voters were “yet to vote,” as absentee voting had not started. During this period, all voters answering the Standard Horse Race question were broken down into 8 categories:
Yes, Not So Strongly
Undecided, Lean Yes
Undecided, Lean No
No, Not So Strongly
These were consolidated into four categories when sample size was low:
Once early voting started, the following categories were used at different times to voters’ answers to the Standard Horse Race question.
The vast majority of the charts in this report use the Combined Proposition 8 Ballot Vote, and therefore use the condensed categories and include those who have Already Voted. This allows us to have consistency between the Standard Horse Race and Be Clear questions, and to show the largest sample sizes available for the Standard Horse Race.
The table below, however, displays all eight categories in the Yet to Vote Standard Horse Race question, and the graphs illustrate the data. It was essential to provide this information in this much detail because this data set shows that the undecided and refused voters, who are combined in the condensed categories, actually show strikingly different trends, especially in early October. At that time, a spike in the undecided/refused number turns out to be entirely due to refusal to answer the question and not to an increase in the number of undecided voters. At the same time, No on 8’s softest supporters—those who described themselves as undecided but leaning toward No—were also cut in half.
Lake’s Standard Horserace for Voters Yet To Vote, Showing Degree of Vote Certainty
This data is most easily comprehensible when seen on the graph below.
Lake Research’s Be Clear Question
After asking the Standard Horse Race question—using the actual ballot language—Lake Research asked a question to clarify the voter’s actual position on the issue of same-sex marriage. Lake asked:
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?
This gave voters a chance to clarify their intention, whether they answered the Standard Horse Race question yes or no. In plain language, this much shorter, clearer question asked voters directly whether they preferred to “eliminate” or “not eliminate” same-sex marriage. This question therefore helped illuminate any voter confusion that might have resulted, eg, from the fact that those opposed to same-sex marriage had to vote “yes on 8 to achieve their policy objection; and that those in favor or same-sex marriage had to vote no.
When the “Be Clear” question is displayed in a table in this appendix, the “Eliminate,” “Don’t Eliminate” and “Undecided” responses are represented as “Y,” “N,” and “U” respectively, since “Eliminate” corresponds with voting “Yes” on the ban. Where space permits—on graphs, for example—the categories are written out in full.
For a full discussion of the differences between the Standard Horse Race question and the Be Clear question, see Finding 2.
Using Lake Research Data to Analyze Voter Confusion
With its Standard Horse Race and Be Clear clarification questions, Lake Research created a set of data that helps show how ballot confusion affected Prop 8. Lake Research looked at voters who had responded to the Standard Horse Race and the Be Clear questions, excluding undecided voters and refusals. The result was four categories:
The data is graphed below in four different ways. The first two graphs show Yes on 8 and No on 8 voters, broken down into right-way and wrong-way voters. The last two graphs show voters who want to eliminate and not eliminate same-sex marriage respectively, again broken down into Yes on 8 and No on 8 voters.
The prevalence of each of these categories made it possible to calculate how much wrong-way voting occurred on each side throughout the campaign, and how much the final vote was affected by wrong-way voting and ballot confusion.
Lake’s Wrong-Way Voting Data
These numbers are represented in the graphs in this section, because they most clearly represent how much right- and wrong-way voters comprised Yes and No on 8’s support over time in the campaign; and how much right- and wrong-way voters comprised supporters and opponents of same-sex marriage.
These figures were calculated from the raw data in the table below.
Lake’s Raw Data Used to Calculate Wrong-Way Voting Numbers
The data in this table is taken from Lake’s polling, and was used to calculate wrong-way voting numbers. The Yes and No on 8 Voters columns show the Standard Horse Race and the Anti– and Pro–Same-Sex Marriage columns show the “Be Clear” question, but only for voters who answered both questions. The “% of Anti Voting Wrong-Way” and “% of Pro Voting Wrong-Way” columns show the percentage of people for each side whose answer to the Standard Horserace differed from their stated vote intention in the “Be Clear” question.
This data was used to calculate the data in the preceding table, which is displayed in the graphs below.
Lake Research’s Standard Horse Race and “Be Clear” Question for Men and Women
Lake Research collected comprehensive demographic data, and broke down the results of the Standard Horse Race question by several of these factors. Here, the Standard Horse Race question is broken down by gender. This data shows that men and women reacted quite differently to various campaign events and showed vastly different levels of support for same-sex marriage at different times during the campaign. These trends within voter subgroups are otherwise obscured in the Standard Horse Race chart for all voters.
The data is presented below in two different ways. The first two graphs compare the Standard Horse Race and the Be Clear questions for men and women separately. Following those, the next two charts compare men and women side-by-side (on the same chart) for both the Standard Horse Race and the Be Clear questions.
For an analysis and discussion of this data, see Finding 1.
Lake’s Polling Broken Down by Gender
This data is displayed in graphs below, for ease of viewing.
Lake’s Standard Horse Race for Men and Women, Broken Down by College Attendance
This data is presented and discussed in the charts below.
Lake’s “Be Clear” Question for Men and Women, Broken Down by College Attendance
This data is presented and discussed in the charts below.
David Binder Research’s Standard Horse Race
David Binder Research conducted daily tracking polls for the No on 8 campaign from October 18 to November 3, the day before Election Day. Binder also conducted one benchmark poll on September 2–4. Binder’s daily tracking polls thus continued after Lake’s ended but started toward the end of the campaign.
In the form the Mentoring Project first received, Binder’s polling was not combined into a rolling average. In order to directly compare Binder’s polling to Lake’s, the Mentoring Project combined Binder’s polling into a rolling average, resulting in the numbers to the right.
Binder’s charts are most easily placed into context when viewed alongside Lake’s charts, as displayed below. Lake’s polls are featured more frequently throughout the report, and the graph below demonstrates why: David Binder Research simply polled for a shorter amount of time than Lake, and thus obtained less data. In addition, the time period during which Binder polled spanned less movement in voter opinion than the period of time when Binder did not poll.
Lake Research & David Binder Research: Latino Voters
Latino voters make up a significant part of California’s voting population; as a result, both the Yes and No on 8 campaigns specifically targeted Latino populations with targeted ad buys and media efforts.
Both Lake and Binder’s polls analyzed voter data by ethnicity and other demographic factors. Lake Research has data for Latino voters for both the Standard Horse Race and Be Clear questions. Where possible, both are presented on the same chart to ease comparison and to provide context. Data is also available from Lake’s polling on Latino voters by gender; this too is presented below. Lake’s breakouts of Latino voters by age, however, result in sample size so small that charts would not be useful for meaningful analysis, and so they are not included. The same difficulty prevented the development of charts on African-American and Asian voters. Regular or periodic oversampling of African-American and/or Asian voters in polling for future campaigns could eliminate this problem.
David Binder Research’s polling of Latino voters is presented alongside Lake’s Standard Horse Race question, again for context.
Some of the charts below illuminate trends for which there is no current explanation. For example, the Lake Standard Horse Race Latino vs White margins shows plunging margins— with No on 8 improving among Latino voters—from October 12 through 17. At the time of this writing, the author does not know why that occurred, what campaign event to which it might be linked, and why the trend then reversed itself.
For a full discussion of this data and its implications, see Finding 1.
Lake and Binder’s Standard Horse Race and Lake’s “Be Clear” Question for Latino Voters
This data is displayed in a variety of charts below.
Lake’s Standard Horserace for Latino Voters Broken Down by Age
This data is displayed in charts below; however, the sample size is low enough that extreme caution should be exercised when drawing conclusions.
Lake Research’s Polling on the “Thorons” ad
Lake’s polling allows us to analyze the effectiveness of the No on 8 campaign’s “Thorons” ad in two ways. (For a transcript and media buy information for this ad, see Appendix E) First, one set of questions gauged memorability. On some days, Lake Research asked open-ended questions about what voters remembered hearing from both the Yes and the No campaigns. Some people recalled seeing the “Thorons” ad. The question was phrased:
Have you seen or heard anything on television, in the mail, on the phone, in the newspaper, or online urging you to vote NO on Proposition 8?
What do you recall hearing or seeing about voting NO on Prop 8?
The table to the right shows the percentage of voters, broken down on the Standard Horse Race question into Yes, No, and Undecided, who recall seeing an “ad about parents wanting their gay daughter treated the same.” The graph of these answers is displayed above.
Second, a different set of questions specifically asked about persuasiveness. Lake asked two questions about this aspect of “Thorons”; they are displayed below. The first, with the same language on all dates listed, asked whether voters recalled an ad matching its description. The second tried to ascertain how “Thorons” affected people’s votes, and was asked in two different ways on different days.
The first day the “Thorons”-related questions were asked, they read together:
Let me ask you something else. Have you seen on television recently a commercial with two older Californians talking about their gay daughter and asking you to vote No on Proposition 8?
[If Yes] And did that make you feel more favorable or less favorable about same-sex marriage?
More favorable/Less favorable/(No difference)/(Don’t know)
On all other days Lake polled on the “Thorons” ad, the second question was different:
Have you seen on television recently a commercial with two older Californians talking about their gay daughter and asking you to vote No on Proposition 8?
[If Yes] And did that make you more or less likely to oppose Prop 8?
More likely/Less likely/(No difference)/(Don’t know)
The latter set frames the second question differently, not asking whether voters were persuaded on same-sex marriage, but rather if their view of the proposition itself was affected. This parallels the difference between the Standard Horse Race question and the Be Clear question. Though the one data point from the earlier set is not definitive, it shows a more pessimistic measurement of effectiveness for “Thorons” on the issue of same-sex marriage. The question asking about “Thorons” and voters’ decision how to vote on Prop 8 suggest some level of effectiveness.
In the tables and graphs below, the first question and the later questions are shown in the same charts. Answers for and against same-sex marriage are displayed in the same data series (more favorable to same-sex marriage with more likely to oppose Prop 8, etc.). This way, similar but nonidentical data sets can be compared and contrasted.
Lake Research’s Questions About the “Thorons” Ad.
The data presented in these tables is presented in the following graphs.
Lake Research Polling on Yes on 8’s Messaging
Lake Research never tested the opposition’s ads, once they were on the air, as thoroughly as they tested the “Thorons” ad, where they asked specific questions about the commercial and how it affected voters’ feelings about Prop 8. Lake did, however, ask open-ended questions about what voters saw and heard from the Yes on 8 campaign. The questions used the following wording:
10. Have you seen or heard anything on television, in the mail, on the phone, in the newspaper, or online urging you to vote YES on Proposition 8?
11. IF YES ABOVE IN Q10: What do you recall hearing or seeing about voting YES on Prop 8?
PROBE: Ask respondent to be as specific as possible.
The interviewer recorded each voter’s first answer to the question, and similar answers were grouped together (for instance, “Gavin Newsom,” or “an ad featuring the mayor of San Francisco,” or any other answer that referred to this particular Yes on 8 ad were put together). With this data, it is possible to get an idea of which Yes on 8 messages were at the forefront of voters’ minds at different points during the most competitive period of the campaign.
Lake Research’s Yes on 8 Message Frequencies for Certain Messages
The information above is most easily understood in the charts on the next pages.
Lake Research’s Polling on Parents
Parents were the obvious target of Yes on 8’s message that children would be taught about same-sex marriage in schools unless Prop 8 passed, a message which was spearheaded by Yes’s “Princes” ad and repeated over the course of the campaign in various forms. To examine how vulnerable this group of voters was to the anti-gay campaign’s messaging, this report examines a number of sets of data from Lake’s polling and elsewhere, to see how parents and other groups of voters actually reacted when the ads were broadcast.
Lake’s polling allows us to look at parents in a few different ways. Do determine parental status, Lake asked respondents “Do you have any children under the age of 18 living at home with you?” This simple question divides voters into two groups, which Lake refers to as “Parents” and “Childless”, but the question has one major limitation: parents with grown children or children living elsewhere are not included, and are indistinguishable both from young adults who do not have children and older voters who have remained childless. However, since we are specifically trying to examine the effects of the “Princes” ad and its surrounding message on the voters it ostensibly targeted—parents with children in school—Lake’s data’s definition of “Parent” and “Childless” breaks voters up into two groups which are perfect for testing our hypothesis.
The first set of charts in this section examines both the Standard Horse Race and the “Be Clear” Question, detailed earlier in this appendix, for both of these categories: “Parents,” or respondents who have one or more children under 18 living at home; and “Childless Voters,” who have no children under 18 living at home. From here on in this appendix, when the terms “Parents” and “Childless” are used, they are used as short-hand to refer to the specific criteria Lake used to create this dataset.
Younger parents and Democratic and Independent voters were ostensibly more supportive of same-sex marriage, part of No on 8’s “base.” To learn about these voters’ vulnerability to the “kids” message, Lake Research ran additional crosstabs after the election, which gives data for the subset of the “Parents” group who are under 45 and identify as Democratic or Independent. This category is referred in short-hand as “Young Parents.”
In summary, the data indicate that “Parents”—voters with a child under 18 living at home—were highly swayed by Yes on 8’s “kids” messaging, especially the Princes ad. Young arents were initially opposed to Prop 8, but showed as large a drop in support as parents at large, switching sides to support the ban
The tables and graphs below illustrate the data in further detail. For further discussion about the movement of parents, see Finding 1. For a discussion of the ads that were most effective in moving this group of voters, see Finding 2.
Lake’s Standard Horse Race for “Parents” and “Childless” Voters
Note that “Fathers” and “Mothers” are male and female voters who answered “Yes” to the question “Do you have any children under the age of 18 living at home with you?”
This data is displayed over the subsequent charts.
Lake’s “Be Clear” Question for “Parents” and “Childless” Voters