This is a slightly different post – It’s a guest post by my friend Russell.
Hello. My name is Russell and I am a guest writer for Kim today. Like many Australians Kim and I were anticipating the results of the Same Sex Marriage (SSM) Survey on Wednesday 15 November. We were both (as I imagine many people were) were interested in how the demographics voted. The Australian Bureau of Statistics (ABS) released results by electorate. Participation rates were also released by gender and age group. We were very keen to know how and why people voted.
As a student of Data Analytics, I had an academic curiosity of this too. The ABS recently released its 2016 Census results. I thought I would try to see if there was any relationship between certain demographics and how their population by electorate voted. I had some findings that were mostly interesting, but also unsurprising. It’s these results I shared with Kim where she asked me to provide her with some content for her blog to share with her followers. I broke down the relationships of how electorates voted into age, income, field of study, occupation, and gender.
Firstly, I want to point out the limitations of this analysis. For the statistic literate among us, I used Pearson’s Coefficient to find the relationship between variables and the vote. The vote was aggregated by electorate, so the power of the result won’t have the same granularity as if I had raw data on how each person voted and their biographical details (which would be unethical). This was a casual analysis and so lacks academic rigour. Nonetheless, a generalised commentary is what I am providing today.
With that business complete, here are my findings.
This bar graph simply shows all variables of all Electorates and their relationship to a vote. Generally, the closer to ±1, the stronger the relationship. I would suggest a correlation greater than 0.5 (or less than -0.5) is an indicator of a good relationship in this data. Don’t worry if you can’t read it – I break it down below.
<Kim comment: For those that aren’t numbers people basically if the line is above 0.5 or -0.5 it means that it’s likely that the relationship between the data is good. Like me and chocolate. So look for the long lines for points of most interest.>
The No Vote was pretty strong for low income earners, but as income increased, the relationship with No voting decreases and even reverses.
As you can see, for people on $65k+ a year, there was a good positive relationship in voting Yes, with the exception of those on $156k+, where the positive relationship was not as strong. Lower income earners tended to vote No, although not as strongly.
In general, age did not play a factor in how an electorate voted. The exception is the 50-59 age group and people younger than 19 where there was a positive relationship with voting No. The vote was not available to children and most adolescents, but their parents were. If I had to hypothesise why, I would suggest the No Campaign’s conflation of SSM and the transgender school’s policy resource programme, Safe Schools, may have had an impact on this demographic. A quick correlation between children and household income would suggest economics was not a contributing factor in this regard.
Field of non-school education mostly showed those with non-school qualifications tended to vote Yes. There was a very strong positive relationship between those whose non-school education was “Not Applicable” (e.g. High School Certificate) and voting No. Those who studied Society and Culture had the strongest positive relationship with voting Yes (Liberal Arts voted for a Liberal position? I’m shocked).
There are about 50 generic fields for job roles. I won’t bore you all with each, so will be equally generic. For the most part, there were only a few standout professions that tended to vote one way or another. Education Professionals had the strongest positive relationship with voting Yes, while Road and Rail Drivers had the strongest positive relationship with voting No. The division has an element of Blue vs. White-Collar workers, but it would be wrong to simplify it that far as there are a few jobs that had very little relationship between votes or even Blue-Collar roles tending to vote Yes and vice versa.
Finally, I find very little positive relationship between gender and how they voted. Men tended to vote No more, but this relationship is fairly weak. Age is useless to visualise because it’s so weak – so no graph for you!
If I could paint a picture of how Australia voted, it follows a typical narrative projected by many commentators. Change vs. Status Quo. The SSM Survey appeared to correspond to how people see themselves in the world and their anxiety of the future (and for their children). There are many variables to analyse, like faith, ethnicity, and effectiveness of campaign strategies. My analysis only covers a portion of the myriad of correlations of how people voted. There’s a danger of going down the rabbit hole of finding that predictor of how people voted from faith-based voting, to how dog and cat owners voted, don’t go down that hole.
I hope you enjoyed this perspective.