In response to Washington Post polling director Jon Cohen's arguments against automated polling, pollster Mark Blumenthal takes issue with the notion that robopolls aren't as accurate: according to some reliable studies, Blumenthal says, robopolls perform about as well as those in which people ask the questions. (He cites analyses by the National Council on Public Polls, the American Association for Public Opinion Research's Ad Hoc Committee on Presidential Primary Polling for 2008, and others).
Those studies only measured the so-called "horse race" numbers--who people are going to vote for--and not broader issues like what voters care about, which Blumenthal acknowledges. Robopolls usually don't go into as much depth, because they're shorter, because people run out of patience more quickly when talking to a computer. (Robopolls may actually have an advantage, if people feel freer to tell the truth to a computer, Blumenthal suggests.)
A broader issue about the value of automated polls, and their horse-race-only data, is this: as long as people are interested in the horse race, the value of the "horse race" polling of robopolls isn't diminishing. Since the interest is there, Blumenthal suggests, it might be time to stop knocking robopolls for what they are.