Cellphones: Health Surveys the Canary in the Coal Mine?
David Brown's article in yesterday's Post (and chat this morning) assesses the difficulty health researchers face on account of the growing population of adults who are reachable only by cellular telephone. Similar thorny questions embroiled the political research industry during the 2008 election cycle. But an analysis of data collected conducted as Barack Obama and John McCain battled for the presidency finds little, if any, skew - although it remains far from a settled question.
To avoid missing potential voters, the Washington Post-ABC News tracking poll conducted in the weeks leading up to Election Day included interviews with those who could be reached only on cellphones. Looking at our own results and the results of the network exit polls for cell-only voters, we found the inclusion of cell-only respondents makes a small difference in the results, with the key word being "small."
Aggregating data from the entire tracking poll, a landline-only sample results in a 51 percent Obama, 45 percent McCain split on the presidential vote question. By adding cellphone only respondents, the same question yields 53 percent for Obama, 44 percent for McCain. That's a difference on the margin - a six-point versus a nine-point lead - but not a statistically significant one: The point estimates for both candidates are not really different, statistically speaking, when calculated using only landline interviews from the result when cell-only voters were included. Further, adjust the results among landline-only respondents so that the composition by age matches the one found in the network exit poll (i.e. pushing 18-29 year olds up to 18 percent of the electorate), and you wind up with a split of 53 percent Obama, 44 percent McCain. This is why weighting landline samples by age has been such an effective substitute for cellphone interviews in many surveys.
The technique is less helpful in health surveys because of their broader reach. When surveying adults, the slice of 18-29 year olds missed by not interviewing on cellphones is far greater, and far different in behavior, than the one missed when the target is voters.
Results from Election Day voters interviewed for the network exit polls generally conform with these results, showing similarly small differences.
But the exit poll also revealed a steep increase in the number of cell-only voters, an increase that matters, because when it comes to interviewing by cellphone, the problems seen in health surveys now may soon infect political research that ignores this population.
For the view from researchers who included respondents who are reached on a cell phone but are not cellphone only, check out the Pew Research Center's report on its pre-election surveys here..
The comments to this entry are closed.