The election may be over, but opinion polls persist. But can public opinion really be measured? Yes, according to UC Davis statistics professor Jessica Utts -- depending on who and what is asked. Key elements to a public opinion poll are a random sample, a sufficiently large sample and neutral wording. One famous "upset" referred to this fall was the 1948 polling by Gallup, Roper and Hearst newspapers, all predicting Dewey would defeat Truman by 5 percent or more of the vote. Truman's victory revealed the inadequacies of a polling method called "quota sampling," says Utts, author of "Seeing Through Statistics," a book about using statistics in everyday life. Another thing that can go wrong is the wording of questions. For example, people in California tended to support "affirmative action" but not "preferential treatment" --opposite reactions to differently worded questions with similar meaning. To assess the value of a poll, "you should find out who was asked, who responded, exactly what was asked, how were respondents contacted, when were respondents contacted, were all results reported, and how many people were asked," Utts recommends.