Political pollsters are having a difficult time receiving responses from residents due to diminishing landline phone use and increasing cell phone use.
Brian F. Schaffner, an associate professor in the Political Science Department at the University of Massachusetts, and Stephen Ansolabehere, a professor in the Department of Political Science at Massachusetts Institute of Technology, found in a recent study that “one in five households rely solely on cell phones.”
Schaffner estimates that between 35 and 40 percent of call recipients are difficult to reach.
The traditional methods of conducting political polls and surveys to the public needs modification to keep up with the population’s changing habits, say the two researchers. Responses from residents are more difficult to receive since the shift from family home phone usage is increasingly moving toward mobilized individual cell phones. Before the age of cell phones, residents could typically be reliably reached on their home phones. Now that people generally have multiple phone lines, it is more difficult to get an accurate unbiased sample from a population.
“People with land lines tend to be older, they are more likely to have families and children, and they tend to move around less. And because of these characteristics, they also tend to be more Republican than people with only cell phones,” Schaffner said. This produces results with Republican bias and excludes those younger, more Democratic residents.
According to Schaffner, Survey USA and Rasmussen Reports are both large opinion research organizations that don’t include cell phones in their contacts.
The prominent survey-conducting firm www.surveyusa.com states on its website that “no other public opinion pollster has the breadth and depth of Survey USA.” The organization has completed some 24,000 research projects and interviewed 36 million Americans.
SurveyUSA ensures viewers that “the probable residential phone numbers that SSI sells to SurveyUSA are just that: phone numbers. They do not come with names, addresses, skin color, income, education, or any other characteristics. We don’t know anything about you before we start asking questions.” While the phone numbers themselves are unbiased and completely random, they are not an accurate representation of the population as a whole, since these numbers are solely based on residents that have land-line numbers, according to the two researchers.
The solution to ensuring that political polls are accurate seems to be elusive. Schaffner notes that pollsters have tried calling both land-line and cell phones, but that method turns out to be very costly and still proves shoddy at retrieving responses. He also says pollsters try to utilize Internet surveys, but they aren’t always effective, considering some do not have access to computers.
Because there is no feasible solution, Schaffenr said “pollsters do work to try to adjust those samples after the fact to make them look more representative.”
On adjusting the data he explains, “The techniques require some expertise in statistics, but the basic notion is that if you have fewer of some group [younger demographics] in your sample than you should have, you weight the data so that each young person who is in your sample counts more to make up for the shortfall.”
Nancy Pierce can be reached at [email protected].
