There’s a lot to work through as we try to unpick what happened, so we’ll step this out in a number of posts, beginning with the basic demographics of who voted for who.
There’s been some lazy census data by electorate analysis done in the rush to explain what happened, but the reality is those are misleading (and damaging, if you consider it gave a whole heap of people licence to bash conservative voters as old, dumb and poor). Age, gender, marriage, kids, income, and education are all poor predictors of voter behaviour. Each one varies only a little bit, but not enough to explain what happened. Knock yourself out with all these pretty pictures… but don’t be surprised if you get to the end and are a little disappointed.
Note: don’t over-read the UAP and PHON figures, remember the number of those voters in the sample is pretty small. Centre Alliance profiled pretty balanced on everything, and there’s even less of them, so we’ve folded them into the ‘other minor’ group.
You’ll also note we’ve ditched the 2PP. As argued last week, it doesn’t work anymore. The Coalition/ALP/Other coding didn’t work very well either (despite my honest belief it would be the better option) once we started working through the figures – so we’ll use the incumbent v others coding as an indicator of who voted for change.
A NOTE ON METHODOLOGY
The Voter Choice Project is a Columbia Panel Study, involving multiple waves of surveys delivered to the same people. The methodology is focused on understanding change in the social context, and involves obtaining qualitative and quantitative information fused in a single survey tool, analysed qualitatively and quantitatively, and with reference to other observations, and to social and historical context.
Participants were recruited via a combination of self-selection, snowball recruitment, and targeted advertising to specific target groups (predominantly on Facebook) to achieve a well balanced sample. Registration for the panel was locked when the election was called both to prevent stacking by partisan supporters, and to ensure multiple surveys were collected from each participant. The panel included 2482 verified voters when the election was called: Coalition voters and women under 30 were slightly under-represented but well within acceptable limits; all 151 electorates were represented with a proportional balance across the States and Territories. The study began in June 2018 with monthly waves, additional surveys for the Super Saturday and Wentworth by-elections, an additional snap survey when Malcolm Turnbull was dumped as Prime Minister, and weekly surveys during the formal campaign. All surveys were administered online, with a 4 day window for weekly surveys and a 10 day window for completion of the monthly waves from October onwards (previously shorter windows proved susceptible to being skewed by outrage reactions to current events).
All earlier numbers on the blog were weighted by age, gender and reported 2016 vote; the below and following numbers are weighted by age, gender and actual 2019 vote. 1418 verified participants completed the Weekly 5 (Exit Poll) survey and indicated their actual vote. Age, gender and electorate of all 1418 participants is recorded. 968 of those had completed the first survey, which collected the other important demographic information detailed below.
The only interesting part in the age figures is that more than half of the Greens’ vote is under 30, other minor party voters are also a younger on balance, and UAP and PHON voters are a bit older. But every party got votes from all ages. For the record, our oldest participants are in their 90’s. We do have a slight male bias in the panel, mainly due to the difficulty of getting women under 30 to participate rather than any over-representation of men. We did have some jokers (mainly older, minor conservative party voting men) who objected to there being an option of a gender other than male or female; where it was indicated by comment they were recoded, but that wasn’t always possible, so don’t read too much into the ‘other minor’ non-binary figure being high. Men are significantly more likely to vote for minor conservative parties, women slightly more likely to Greens or Independents.
The marriage stats largely follow the age stats, with the exception of other minor party voters who are more likely to be married (in contrast to the very well balanced age groups). The ‘other’ for children includes people who have foster children, care for grandchildren, and others who noted their children are long grown up. Greens’ voters are the only ones less likely to have children, which again tracks with the age stats.
Income is extraordinarily well spread across all parties. Although I did get a chuckle from Labor having more highest earners than the Coalition – that goes against the narrative. Notably, PHON voters are poorer, but caution that there is only a small number of PHON voters, so don’t over-read it.
Tracking with income, PHON voters are likely to be less educated. Greens, other minor and independent voters are more likely than not to have some university education, and there are very few PhDs voting for the Coalition.
See? Not much there. Sorry. In the next post we’ll look at what did happen.