In the first post I went through what the long text comments or qualitative information told us happened in the Wentworth By-election; in this post we’ll crunch the numbers for a different view of what happened. It’s the same story: Licia Heath and Tim Murray got beaten badly by a late shift to Kerryn Phelps, but there’s a number of additional insights (and lots of pretty graphs).
SAMPLE
391 people (214 men, 175 women and 2 non-binary gender identity people) responded to either Survey 3 of the panel study or the ‘exit poll’ survey, both conducted between the evening of Thursday October 18 and Sunday October 21. 251 participated in both a previous survey and this final survey, enabling analysis of vote shift. Numbers have been weighted to age, gender, and the current 2018 by-election results at the time of writing although we do not expect enough of a shift from here to make a substantive difference to the results. The sample is much smaller than we would like (resulting in about a 5 point MOE), but is reasonably well balanced, and the panel aspect is enormously valuable.
As you can see above we continue to have issues recruiting younger women to participate in all our surveys, as we do stable Liberal voters. These issues are corrected by weighting, but we continue to work on ways to engage these parts of the community to participate in our main study (sign up here).
MOVEMENT OF VOTES
A third of respondents who participated in the ‘exit wave’ survey and a previous survey changed their vote in the last week. That’s not a typo, or rounding – it’s literally 33%. 37% were voting Liberal and stayed there, 10% voting Labor stayed there, and only 2% of the vote stayed with Licia Heath. Phelps retained 20% of the vote, and mugged everyone, taking at least a quarter of the vote off everyone else except Dave Sharma.
The below is our vote retention graph. Along the diagonal is the percentage of vote each candidate has retained since the last surveys – the quick poll conducted between October 6 and 9, and Survey 2 of the panel study conducted between October 10 and 14. Read across the rows to see where they have gained votes from, and down the columns to see where they have lost votes to. The percentage figures are percentage of the early October vote for that column, and the colour indicates who they ultimately voted for.
The why is best explained in post 1, but you can see that while the Greens lost (net) 17% of their voters to Phelps (the 4% of Phelp’s vote coming back to them equating 10% of their vote), they gained more than that off Tim Murray – thus their polling day vote was a little higher than where the polling had been a week prior. By comparison, both Heath and particularly Murray lost huge numbers to Phelps, and didn’t get much back from anywhere, leading to that unexpectedly low result for both of them. Sharma got his vote up from the high 30s to the 40’s entirely from minor party (mainly Liberal Democrats) and a few undecided voters – the exchange with Phelps is actually pretty even. (Note this paragraph has been updated to correct an error: previously said the Greens net loss to Phelps was 23%.)
To look at these figures another way, this graph shows the same thing, but the heat map and percentages are as a share of the total vote (as opposed to per candidate). I don’t normally show this as it’s not as illuminating on what’s going on for an individual candidate’s vote – it’s certainly not as demonstrative of extent to which Phelps cleaned up, but it’s better for understanding what the swapping of votes mean in real terms.
EVALUATING THE QUICK POLL
This vote movement as a share of total vote graph is the best to use to evaluate the performance of the experimental quick poll we conducted. The poll had a number of untried features and unproven methods as we search for better ways to poll. Initially on election night, while some were congratulating us on picking a 55-45 result, I was in despair at how it had been so wrong (as I knew the pre-poll and postals would bring it back a couple of points, and the Heath and Murray votes were so off the mark). This shows the poll wasn’t that wrong, the vote shifted – and we need to change the way we talk about polls. It is simply not reasonable to expect a poll conducted from the 6th to the 9th to pick up a shift that happens on the 15th.
The quick poll numbers as published were:
Sharma 38.8
Phelps 23.5
Murray 17.2
Heath 7.4
Wy Kanak 5.9
Others 6.5
Not sure 0.8.
The numbers in the above graph are a different sample, drawn from the two early October surveys, filtered by those who completed the last survey, weighted to the 2018 by-election results – instead of the 2016 vote, and without the additional suburb and vote intent weightings. If you add up the numbers in the columns, you essentially get a new set of figures for where the electorate was at a week before the election… which is:
Sharma 37.7
Phelps 22.8
Murray 16.9
Wy Kanak 8.45
Heath 5.9
Others 6.6
Not sure 1.6
That’s very close to what we had, albeit very different from the result, and with a much higher margin of error (given the sample of 251 instead of 700+). The Greens and Heath numbers are a little too different to be comfortable with. What we gain from this is both improvements to our methods, and a very clear message that a week out is too far away from election for a poll to be considered ‘predictive’. We’ll work on it some more.
HOW TO VOTES AND BOOTH FACTORS
To give further weight to the evidence of a strategic vote, we asked whether people used a how to vote card when voting. This is pretty low by most estimates, but there isn’t good data on it to compare to. It’s lower than the Super Saturday by-elections.
The Non-Party HTV’s were coded from the ‘other’ responses, and were mostly people saying they used GetUp’s HTV. There may have been a great deal more than that, as it wasn’t prompted, and the question didn’t specify that the HTV had to be a party’s HTV.
We also asked a question suggested by Antony Green, which was whether their party was handing out at the booth – theory being that one of the main reasons people don’t follow HTVs is their preferred party isn’t at the booth to give them one. Not the case so much in the hotly contested Wentworth by-election with 16 candidates making HTV distribution a campaign priority, although 8% still said their candidate’s HTV wasn’t there. The don’t know/others were people saying they didn’t look for one, or deliberately avoided the gauntlet so didn’t know.
Seeing candidates at the booth – noted as a significant impact on final vote decision – was also quite high. Although there were a lot of candidates for people to see, they clearly hustled throughout the entire day. The ‘others’ were people volunteering they had seen other figures such as Labor’s Anthony Albanese and Liberal Democrats leader David Leyonhjelm.
There was a clear divergence between respondents who very clearly over the campaign, with many negative comments about the volume of Liberal Party advertising and mail in particular, and others that wanted more information. Some were specific that they wanted more information about the ‘other’ candidates, and were frustrated with coverage being so focussed on just the front runners, and some wanted more ‘meet the candidates’ forums (with all candidates attending, please). The vast majority, however, did feel they had enough information to make an informed vote. They also didn’t wait long to vote. So the by-election passed the ‘good democratic process’ test.
MINORITY VOTE
We asked people to identify if they identified as either of the two significant minorities frequently discussed in this campaign – Jewish or LGBTIQ+. The numbers are fairly small, so these numbers should be read with caution, however there is a shift from the first panel survey which had a much more balanced spread of both minorities across the voting options (although the numbers aren’t directly comparable, and the numbers are far too small to publish a vote movement graph). As the campaign neared its end Jews went to the Libs in numbers, LGBTIQ+ to Phelps, but not as strongly.
One for you, one for me…
MMM… SAUSAGE.
Lastly, this is the most important question of all the questions in all the Voter Choice Project surveys.
Wentworth, you can do better. Particularly the 40% denied because there was no sausage on offer – such a great fundraising opportunity lost. There’s quite a few ‘others’ here… a number of people got democracy pancakes! I like it. There was also cupcakes, of which we heartily approve, and sausage sizzles that either weren’t ready when the polls opened or had run out of sausages, of which we do not approve. And vegetarians, who were clearly not where the cupcakes or pancakes were.
Some asked why I ask this question, and yes, in part it’s because I have fun doing this and want you too as well. But the level of #democracysausage activity is also a proxy for how engaged the broader community is the electoral process. There is method in my mischievousness.
Thank so much to the people of Wentworth for giving us these fantastic insights into a very intense race. I need to thank so many of the candidates too for promoting the study, enabling the decent sized sample and our work to make significant advancements in the last 6 weeks.
Comments are closed.