One of the state’s most respected polls has begun incorporating online surveys for the first time, underscoring the increasing difficulty of relying on telephone questioning.
The Field Poll, which was founded in 1947, started using online surveys to gather voter opinion on nine of the 17 statewide ballot propositions that will appear on the Nov. 8 ballot. The Field Poll conducted the surveys in partnership with the Institute of Governmental Studies at the University of California at Berkeley and YouGov, an international market research firm headquartered in the United Kingdom. Results were released last month.
Mark DiCamillo, director of the Field Poll, said the main reason he used online was because he believed it was the best way to get voters’ opinions on a large number of ballot measures. He wanted to give voters the exact title and wording of the measures and he knew that for telephone surveys, voters will typically only give their opinion for two or three propositions.
When he started doing telephone polling for Field in 1978, the cooperation of people who answered the phone was 65-70 percent. Today, it’s only about 30 percent.
“You never know when they’re tuning out,” he said. “You can’t ask more than a few because you’ll lose the respondent on the phone.”
DiCamillo added that Field Research Corp, which owns the Field Poll, has done online polling in the past.
The polling industry as a whole has had a harder time in recent years reaching people with telephone surveys.
As an example, DiCamillo said that when he started doing telephone polling for Field in 1978, the cooperation of people who answered the phone was 65-70 percent. Today, it’s only about 30 percent. “A lot of the people won’t even pick up the phone,” DiCamillo said. “That 30 percent is of the people we actually reach. The actual response rate of all the numbers you’re dialing that are usable is less than 10 percent.”
The advent of cell phones also made the process more difficult, DiCamillo said. Telephone surveys used to just be on landlines but now less than two-thirds of all Americans have a landline phone at home. Both types of phones have to be surveyed. Because of call screening, response rates have also gone down. “It becomes a concern,” DiCamillo said.
The margin of error in a sample is determined by the formula: 1 divided by the square root of the number of people in the sample.
DiCamillo emphasized that telephone surveys still work but they take a lot more effort.
The main advantage of Internet polling is the cost, said Jack Citrin, director of the Institute of Governmental Studies. With online polling, you eliminate the need for employees to make calls and then call respondents back who aren’t immediately available. Another advantage is that people may be more willing to say things online that they wouldn’t on the phone.
“We know from social media that people are willing to say things online that they wouldn’t say to your face,” Citrin said.
Still, telephone surveys remain popular because they are a method of getting a random sampling from a known population. A polling firm can telephone a random sample of all available phone numbers. Then the polling firm can calculate the sampling error — the error that arises from taking a sample from a population rather than using the whole population.
The margin of error in a sample is determined by the formula: 1 divided by the square root of the number of people in the sample. So for instance, in a telephone survey of 1,000 voters from a field of 18 million California registered voters, the margin of error is 3 percentage points.
It is not possible to calculate the margin of error in an online poll because there is no “sampling frame for the Internet, no way to draw a national sample for which virtually everyone has a chance of being selected,” according to the Pew Research Center’s May report “Evaluating Online Nonprobability Surveys.”
“As many people have emails now as have phones.” — Eric Jaye
The accuracy of an Internet political poll depends on the mix of people in the online panel- whether they accurately reflect the views of the voter base as a whole. “It’s a different methodology,” said DiCamillo of the Field Poll. “You’re not doing a random sample. You’re doing a model of the electorate, bringing in people of all ages, all regions and all parties.”
DiCamillo said he chose to partner with YouGov because it does the best job of putting together online panels. Pew Research Center recently identified YouGov as the most accurate of the nine companies they evaluated. In August, poll rating website FiveThirtyEight reported that YouGov correctly called 93 percent of the races in 707 polls analyzed.
Political consultant Eric Jaye, founder and president of Storefront Media in San Francisco, said it’s difficult to make decisions about polling because all the technologies have a self-selection bias. Online polls are thought to have a greater self-selection bias because people choose to take an online poll and some people are not represented, he said. But that is changing. “As many people have emails now as have phones,” Jaye said. “My 90-year-old father responds to my emails most days better than my 21-year-old daughter.”
Jaye advises clients to use both online and telephone polls to get the best information.
“In general, it’s better to use all the channels,” he said.
Ed’s Note: Click here to see the Pew Research Center’s evaluation of nonprobability polling, and here’s the link to 538’s pollster rating article.