The Field Poll of 830 likely voters found that the share of the electorate backing Proposition 8 fell during the last two months, with 38 percent of those surveyed saying they intend to vote for the measure compared to 42 percent in early July.
Opposition to Proposition 8 increased during the same period, the poll found. A solid majority of the likely voters—55 percent—said they would vote against the same-sex marriage ban, compared to the 51 percent who opposed the initiative in the July poll.
"It's its clear that when Californians learn what Proposition 8 will do, which is eliminate an existing constitutional right from one targeted group of people, they move toward voting 'no,'" said Geoffrey Kors, executive director of the gay rights group Equality California.
Also known as the Marriage Protection Act, Proposition 8 would amend the state constitution to limit marriage to a man and a woman. If passed, it would overturn the California Supreme Court decision that made the state only the second in the U.S. to legalize same-sex marriage.
Field Poll director Mark DiCamillo said the poll showed that Attorney General Jerry Brown's July 25 decision to change the proposed amendment's phrasing to reflect that same-sex marriage was now legal played a role in eroding public support for the measure.
Before the Supreme Court issued its sweeping decision on May 15, the referendum's sponsors qualified it for the November ballot under the title "Limit on Marriage." Its original summary was confined to the exact language that would be inserted into the constitution: "Only marriage between a man and a woman is valid or recognized in California."
Brown's office said that language was no longer a sufficiently accurate description once the court's ruling took effect and rewrote the title and summary to say the proposition would "eliminate the right of same-sex couples to marry." The measure's backers sued, arguing that the new phrasing unnecessarily prejudiced voters against the measure, but a judge sided with the attorney general.
To gauge the effect on voters, the Field Poll read one half of its respondents the original language and the other half the revised version.
Among the 70 percent of likely voters who already were familiar with Proposition 8, the modification appeared to make little difference. Among those who knew about the amendment, 56 percent said they opposed it when they heard the original wording and 53 percent opposed it they were given Brown's revised version.
But among the 30 percent of those surveyed who were not previously aware of the measure, the ballot language appeared to matter. Within that subgroup, 42 percent of the respondents said they were inclined to vote 'no' with the original summary, a share that climbed to 58 percent under the new wording.
"I would hazard a guess that it's the words 'Eliminates the right to do something' that is a stimulus in their making a judgment," DiCamillo said. "Whatever the right is, generally speaking, voters are in favor of individual rights. ... So I do think there is a negative drag in that compared to the original wording for those who are not paying attention."
Jennifer Kerns, a spokeswoman for ProtectMarriage.com, the coalition of religious and social conservative groups that put the measure on the ballot, said the poll results were not surprising.
"It's not surprising given the attorney general's latest attempt to influence the elections. We were expecting it would affect the numbers by a few percentage points," Kerns said.
At the same time, she said the Yes on 8 campaign's internal polling shows voters to be much more evenly divided and the initiative's backers expect support to pick up once they start airing television commercials later this month.
Kerns also disputed the Field Poll's accuracy, noting that in the weeks before California voters considered a gay marriage ban in March 2000, the company found support for it topping out at 53 percent. The measure—one of two marriage laws the Supreme Court overturned as unconstitutional—passed with more than 61 percent of the vote.
Regardless of the wording change's effect on the electorate, DiCamillo said the new numbers do not bode well for the marriage ban.
"The chances of passage are extremely low," he said. "We have polled on more than a hundred ballot initiatives, and history shows that when an initiative starts out behind, it very rarely passes.
"When you couple that with the fact that this is an issue about which people have strong feelings, what that says to me is an initiative like this is less susceptible to wide swings in public opinion."