LA Weekly

June 12-18, 1998

The Poverty of the Polls
Latinos outfox opinion surveys – or vice versa.

Contradicting pollsters and campaign rhetoric, the predicted across-the-board backing for Proposition 227 appeared to vanish suddenly – if indeed it ever existed – on Election Day, when five out of eight Latino voters and half of African-American voters said no to the plan to gut bilingual education, according to an L.A. Times exit poll of more than 5,000 voters.

The June 2 poll result was a stunning reversal of the Times's last pre-election canvass, taken only two weeks earlier, which claimed Latinos backed 227 by an overwhelming margin of 62 percent to 26 percent. That Times poll also found Latino opinion to be moving toward a stronger yes vote in the six weeks since its previous April survey, which (if true) would make a last-minute switch to opposition even more puzzling.

These figures are explained in quite different ways by pollsters and Latino leaders. Mark DiCamillo of the San Francisco–based Field Poll (whose late-May survey at least discerned falling support for the measure among minorities, though it still had Latinos backing it 52 percent to 38 percent) suggests that "maybe they come to judgment later," citing sharp drops in Latino backing for propositions 187 and 209 as those campaigns swung into gear. In the case of 209, he said, the final 30-70 Latino outcome opposing the ban on affirmative action was a complete reversal of 70- 30 support a few months earlier. A reason for the switches on 227, DiCamillo says, might have been last-minute advertising by the "no" forces targeting Latino voters.

Jill Darling Richardson, assistant director at the Times poll, shrugs off the outcome as, "not a big surprise after our experience with 187 and 209." Polling, she added, "is more an art than a science."

In contrast, Antonio Gonzalez of the William C. Velasquez Institute (formerly Southwest Voter Research Institute) charges the pollsters with sloppy methodology and maintains that anyone who bothered to poll an adequate number of Latinos could have forecast the outcome. "They're going to cover it up by saying we decide at the last moment, but it's really their profoundly bad social science." The number of Latinos polled, statistical authorities agree, is far too small to get an accurate picture of opinion at any point, let alone trends. The Times's polls were based on about 1,100 registered voters, and about half of these were considered not likely to vote. Seventeen percent of the registered voters in the poll were Latino, a base so small as to make predictions difficult. The sample drawn by the Field Poll was based on a sample of 714 likely voters.

A poll in January, designed by Loyola Marymount political science professor Fernando Guerra, tends to confirm Gonzalez's point. Guerra's poll of 650 registered Latinos asked half those polled the question as it was summarized on the ballot; the other half also heard that it would "eliminate" bilingual education, a description that Guerra expected opponents to use during the campaign.

The first group supported the proposal 65-23 (much like the Times poll findings); the other half opposed it 64-31 (much like the real outcome). In the long run, says Guerra, the results show that surveys need to simulate the campaign environment.

In the short run, Gonzalez emphasizes, pollster error may have influenced the results of the balloting. Critics of bilingual education made hay of the idea that even Latinos opposed continuing with Spanish-language instruction. Says Gonzalez of the poll-takers: "Due to their sloppiness, they functioned as accomplices to Ron Unz and the pro-227 forces."