But none of the polls has seriously tested what I’m interested in: which of them would voters actually support in a general election?
The polls of the membership (and registered supporters) have looked at which of the candidates those people support.
While the accuracy of the sample of these polls is hugely questionable – and so they might be essentially meaningless [Update: Election Data has made a good case for why these membership polls seem pretty good] I have no problem with the question they ask. They’re trying to find out who’s likely to win the leadership election, and do that as best they can.
So far, though, there haven’t been any particularly useful polls of the public. The recent Mori poll asked people about their current views of the candidates. The trouble is, most people have very little idea about the candidates, so the polls are partly a name-recognition exercise and partly a reflection of the tiny amount of knowledge people currently have.
I’m not interested in what people think about candidates they essentially know nothing about. I want to know how they might vote in 2020 if each of the candidates were leader.
To answer that, I’d like to see a poll that did something like the following:
Firstly, a video test of each of the candidates. Show respondents a 30-45 second clip of each candidate. It’s essential they’re talking about the same policy area and in as similar an environment as possible. A Survation poll did this before the election but was limited by having the candidates talking about different issues. I’d even consider giving a couple of (shorter) clips of each to make it as balanced as possible.
Then, ask candidates who of those four they’d most want to vote for. You could test a series of attributes as well, like who is the strongest, most in touch with ordinary people, willing to take tough decisions and so on – but I’m much more interested in the gut response of who people support since that incorporates all the attributes and weights them according to importance.
This video test would, on its own, be much more useful than the questions based on current knowledge.
I’d go further though, and test how vulnerable the candidates are to the attacks they’d inevitably face as leader. I’d try something like this: