To put it bluntly, I’m not happy with any of the polling out of Wentworth. I launched a quick poll today, which is not part of the by-election study, but is part of my other work in finding a better way to poll.
Don’t know, but Simon Jackman and Luke Mansillo at the University of Sydney’s United States Studies Centre, both of whom I have a great deal of respect for, found that seat level polling in the 2016 federal election was as inaccurate as a poll with one-sixth the reported sample size.
So what kind of sample do you need for a seat level poll to be accurate? Or is it an issue of methodology?
Well, let’s find out.
There are a number of experimental features in the Voter Choice Project, but the point of the VCP is determining the why behind the nation’s votes: and so the focus on recruitment has been more on people who will stay with the panel through to the end than just sheer numbers. The surveys are longer and more detailed, with a lot of validation and probing questions.
This poll has streamlined or removed a number of those features to get it down to an average completion time of 5 minutes. I have made it a requirement that an email address be provided to try and stop candidate’s supporters from answering the poll more than once (although I do want all supporters to take part, and share it with their friends and family in Wentworth, and I will be asking the candidates to share it with their followers on social media).
Analysis will be done at 800, 1200, 2500 – and if I am fortunate to get that many, 5,000, and at each 5,000 increment from there. Then we’ll see how many it takes to get to an honest reflection of the electorate in a seat level poll.
Please help us in this work by encouraging anyone you know in Wentworth to take 5 minutes and do our Quick Poll.