The Robo-Polls vs. Live-Polls Divide

Charlie Cook, one of the nation’s leading polling analysts for years, made this important note in his latest article:

The problem with state polls is that most are in the extraterrestrial category; robo-polls are often all over the map. Aficionados would be well advised to focus on state-level polling offered by NBC/Wall Street Journal/Marist University; CBS/New York Times/Quinnipiac University; and by ABC, CNN, Fox News, and other brand names that specifically use live interviewers calling voters with landlines as well as the 30 to 40 percent of voters (mostly young people and minorities) who have only cell phones. Although poll aggregators have a good technique in averaging data, be advised that a lot of dubious information goes into those averages; it’s wiser to focus on the brand names with the more traditional (and very costly) methodology.

Both Cook and Nate Silver have expressed concern about the accuracy of robo-polls. Some major robo-pollsters include Rasmussen, Public Policy, Gravis Marketing, and ARG, to name a few.

Just to give you an idea of how sharp the difference is, if you average together the latest four polls of Ohio, Obama’s lead is 1.5%. If you use only the latest polls that used live interviews, Obama’s lead is 5%. In Virginia, Obama’s lead when averaging the latest four polls is 0.5%, but if you use only live interview polls, it rises to 2.0%. This will be an interesting thing to evaluate after the election, to see if the live-polls do turn out to be more accurate, and what that means for the future of robo-polling, which is currently on the rise because it is cheaper to do.