Advertisement
YOU ARE HERE: LAT HomeCollections

Q&A POLLING RESEARCH

Tracking Public Opinion Is a Delicate Blend of Science, Art

October 04, 2005|William Nottingham | Times Staff Writer

So how do you really feel about abortion? Immigration, the economy or Iraq? Or the president? Or the governor?

With the fall election season heating up in California and other states, public opinion pollsters will be asking questions on such topics to thousands of people in a drive to pin down the ever-shifting mood of the country.

As Los Angeles Times Poll Director Susan Pinkus explains below, obtaining a valid result involves a keen blend of art and science.

Gallup, Roper, Field and newspapers such as The Times -- to name just a few major polling organizations -- begin with a bank of telephone callers who ask carefully written questions of people scientifically chosen at random. Randomness is a key that allows pollsters to accurately project the opinions of thousands based on a sample of only 1,000 or even hundreds.

The answers are entered into a computer and sorted against various statistical measures. The results are then analyzed in the aggregate and qualified with a mathematical "margin of error" that tells readers the results might vary within that range.

*

Question: What is the value of a public opinion poll?

Answer: It tells us what people are thinking at that particular moment about a particular issue. Polling is just a snapshot in time, and it's not a predictor of the future. It's just basically telling you, if there's a [polling period] of let's say, Oct. 1 to Oct. 7, that's what voters were thinking about those issues during that period of time. We can get a sense of what Americans are feeling about the war in Iraq, for example, or the government response to Hurricane Katrina, or the special election in California.

*

Q: As a professional pollster, how do you react when you hear a politician say something like "Oh, I don't believe in polls"?

A: I laugh. It's probably because they have a result they don't really like. Because I am sure that they all have their own pollsters, they all care about public opinion and they all want to be on the right side of it so that they get the votes the next time they run for election.

*

Q: Websites occasionally ask a "question of the day" and then show how people responded -- is that a poll?

A: It's not a poll because it's not a scientifically designed sample. You can have a website where 10,000 people get on it, saying whether you want John O'Hurley or Kelly Monaco to win "Dancing With the Stars." That's a simple issue and that's a silly issue. But does it matter? No.

But it's different when we're talking about current events or timely issues, like [President] Bush's job approval rating or whether you think abortion should be legal or not. When a website says "OK, click yes if you agree" there could be a systematic or a concerted effort from one major group to answer it multiple times. You don't know if it's children answering it or only adults answering it or whether they've answered it many times.

*

Q: So what makes a poll scientifically accurate?

A: We begin with a "probability sample," which allows every adult 18 or over to be chosen to participate in the survey. We use a random-digit dialing technique that uses computer-generated telephone numbers.

And you only want to speak to one person in a household who is 18 or over. We don't even let the interviewers decide whom to choose because for the most part women answer the phone, so if you didn't balance it out you would have a very one-sided survey of more women answering than men. So even within the household we randomly pick somebody by asking to speak to the person with the most recent birthday.

It's a scientifically produced sample, so that you get representative numbers of men and women. You compare some demographics to census data. A random sample will have the right amount of men and women, the right proportion of blacks, whites, Latinos and Asians.

*

Q: Does it matter what kinds of people are questioned?

A: It depends on what your poll is trying to do. If you're getting close to an election, you want to look at "likely voters" because those are the people who are going to vote. You don't want to look at everybody, because not everybody votes. We just saw that in our last presidential election: Only about 54% of the population voted.

*

Q: How do you determine who is a "likely voter"?

A: We ask the person we're calling about their past voting history, their intention to vote, their interest in voting, are they a citizen, are they registered to vote, whether they are definitely or probably going to vote.

*

Q: Is there any magic to the way a poll question is phrased?

A: Polling is called an art and a science. The science is the sampling and the methodology, and the art is the way you design the survey questionnaire. You have to be very careful of the order in which you ask the questions so that one question doesn't taint the next question.

Advertisement
Los Angeles Times Articles
|
|
|