We’re starting, once again this election cycle, to see some pollsters report results for self-described “definite” or “certain” voters, as well as for registered and, perhaps, likely voters.
maristpoll.marist.edu/…
Two years ago, Iowa pollster extraordinaire Ann Selzer reported in an interview with FiveThirtyEight she only includes definite voters (“they could have said they were probably going to vote… We don’t take that”).
fivethirtyeight.com/…
Monmouth University’s Patrick Murray appeared later on in the same episode. He noted that he had data from six states that included definiteness of voting and other possible markers of voting likelihood (e.g., voting history). As he reported, there was only one state in which using a definite-voter criterion improved the accuracy of the turnout model — Iowa.
In addition, some people who would not consider themselves definite voters may still vote. In fact, Obama was famous for his campaign organizations’ ability to bring out “unlikely voters.”
www.reuters.com/…
Definite-voter models may work occasionally, but I would treat them simply as one piece of data in the mix.