Ward Connerly's ballot initiative to restrict collection of race-based data continues to enjoy more support than opposition among California voters -- but the gap is narrowing, according to a Field Poll released Monday.

Forty-six percent of likely voters said they would vote yes on Proposition 54 if the election were held today, compared with 35 percent who would vote no and 19 percent who are undecided.

Support has fallen since last month, however, when half the voters surveyed gave Connerly's measure a thumbs up and 29 percent said they would vote against it.

Mark DiCamillo, Field Poll spokesman, cautioned against drawing conclusions about Proposition 54's ultimate fate based on early poll numbers. But it can be ominous whenever support for a ballot initiative falls below 50 percent and large numbers of voters are undecided, he said.

"If you're undecided when you enter the ballot box, you're more likely to vote no," DiCamillo said.

Lack of information may be a key reason for voter indecisiveness: Fifty-nine percent of respondents said they know little about Proposition 54.

Connerly's measure would ban state and local government from collecting race-based data except for medical research, federal requirements, legislative authorizations and several other purposes.

Proponents say the initiative would push government toward becoming more colorblind by restricting data that divides people by skin tone. Opponents counter that Proposition 54 would make it harder to identify and correct inequities.

Thomas Burns, a 44-year-old Yuba City resident surveyed for the Field Poll, said government has no business classifying people by race.

"The more the state gets involved and the more the media talks about it, the more it just keeps problems on the fire," he said.

Sharron Wolfe, 65, of Oakhurst in Madera County said she generally objects to race-based statistics but will not vote for Proposition 54 unless adequate exemptions exist for health care data.

Some races can be disproportionately affected by a disease, such as African Americans with sickle-cell anemia, she said.

"If they need it for medical reasons, OK, but just to tag people (by race) is wrong," she said.

Proposition 54 provides an exemption for medical research, but there is substantial disagreement over interpretation. Courts or the Legislature ultimately may decide the issue.

Elena Stern, a spokeswoman for the No on 54 campaign, said the new Field Poll results confirm what she has said for weeks: Voters like the notion of a colorblind society but not Proposition 54 itself.

"At face value, the initiative seems benign, but when people learn about the ramifications and the widespread impacts, they tend to oppose it," Stern said.

Diane Schachterle, coordinator of the Yes on 54 campaign, said she is not disappointed by the poll results in light of "hysterical attacks" in which opponents grossly exaggerate Proposition 54's impact.

"I think you have to put them into context," she said of the poll numbers. "It's still real early. But given the volume of attention so far, we're encouraged that folks are not buying wholesale into misinformation from the opposition and the media in general."

Fifty-five percent of Republicans and 52 percent of nonpartisan voters support Proposition 54, compared with 36 percent of Democrats, the new poll found.

More minority voters oppose Proposition 54 than support it, while the reverse is true among white voters who are not Hispanic, the poll found.

Forty-seven percent of white respondents said they would vote yes; 33 percent said no. By comparison, the numbers for Latinos were 38 percent and 50 percent; for other races, 35 percent and 41 percent.

Among age groups, Proposition 54 is most popular with senior citizens, 51 percent of whom favor it. Its weakest support, 39 percent, comes from voters ages 50 to 64.

Roughly one in four Californians surveyed said the presence of Proposition 54 makes them more inclined to vote in the Oct. 7 recall election.

Seventeen percent of respondents said a candidate's position on the measure will be very important in determining how they vote in the gubernatorial recall.

The Field Poll was based on interviews with 629 registered voters, 448 of whom were deemed likely to cast ballots Oct. 7. The poll's margin of error is plus or minus five percentage points.