And there should be changes to the Elections Act that include requiring pollsters to state who paid for their surveys, the exact wording of their questions and the order in which the questions were asked.
"That stuff is critical for us to know whether we're being manipulated," said Pilon. "Polling is about trying to create the result you want. That's why millions of dollars are spent on it.”
Polls affect the public in various ways, he said. If people feel an election is already decided, they might not bother to vote. If they feel their first choice isn't popular, they might stay home on voting day. There's also the “winner affect” where some will want to vote for the person they feel is most likely to win.
“That's where all this polling crap starts to become very important,” said Pilon.
Pollsters should also state (so say I) whether an introduction or preamble was used at the beginning of their poll or before each question.
Push polling in particular, but even otherwise well-designed polls, will use preamble that can plant a bias in respondents' minds.
For a good discussion of what went right and what went wrong in polling for the 2006 federal election, check out this paper by Bea Vongdouangchanh and Kady O’Malley.
Hands down, the pollster who got the most accurate results during the 2006 election was Nik Nanos, who reported his daily tracking results each evening on CPAC's Prime Time Politics - and is doing it again this election. His projection was within 0.1% of the actual voting results.
UPDATE: See also Impolitical's take on polls.
Recommend this post