Why Ned Flanders would probably vote TurnbulL

You may well be thinking that Christians tend to be richer and more likely to live in rural areas and it may be that which is driving these trends.  That would indeed be a good point.


We can therefore filter the data to account for these factors.  This is done below.  The data is filtered initially by just those living in the capital cities.  I have also broken the data into five charts representing the five quintiles of socio-economic wealth (based on the SEIFA Index of Relative Socio-Economic Disadvantage).  "Wealth 0-20%" is the bottom quintile (the least wealthy or most deprived) and so on.  This means we are only comparing amongst groups of similar wealth or deprivation.  You can click on the image to freeze scrolling and flick through manually. 

Following on from my previous two posts ('The Data Science of Religion' and 'First Preference Insight with 2016 Election Data') several people have asked if we can determine if there were any religious based swings in the voting of the recent 2016 Australian Federal Elections.


The answer is, with some quite sweeping assumptions, yes.  


And there are some interesting findings.  Christian voters skew strongly right to the Coalition.  Curiously this swing is most significant at both extremes of the wealth/deprivation spectrum - the most prosperous and most deprived moving more to the Coalition compared to similar non-Christians.  Green voters tend to have no religion; they are also skewed to the wealthy within this group.  Labour relatively struggles amongst both Christians (especially within their key group of the most deprived) and those with no religion.  Labour is thus more effective amongst other religions (which ties to the preference to Labour amongst immigrants as we saw in my previous blog).  Meanwhile One Nation and the Xenophon Team are most effective amongst the more deprived and less religious. 


Ideally this sort of analysis would be done in individualised survey data - be it Election Day exit polls which ask each individual's religion, or even pre-election polling or lifestyle surveys.  These results would directly match between an individuals voting preference and their religious affiliation with great accuracy - only limited by the size and representativeness of the survey.


The alternative method I have used here is bundling together aggregated data.  We are using polling booth data which is aggregated and consequently assumes people mostly vote at their nearest voting location.  We are also using religion data which is grouped up to SA1 level (which is about 500 people on average).  This data is also limited by being five years old and asks only religious affiliation, not if you are indeed practicing that religion (and to what capacity).  You may wonder what can possibly be derived from this odd mix of data.  As far as precise findings and accurate estimations, not much.  But as far as broad observations and skews go - quite a lot. 


Below we can see the voting decisions based on the percentage of Christians in the local SA1 area.  On the left we see the voting mix of those areas of few Christians, ranging to those areas on the right with high density of Christians.  Note that the various parts of the Coalition have been coloured shades of blue for visualisation purposes.  Immediately we can see a left-leaning swing for non-Christians and right-learning for Christians.  Votes for Labour erode as the percentage of Christians increase, and even more so for the Greens.  The Coalition is the clear beneficiary of this along with, unsurprisingly, the Christian Democrats.

Using election data to show the Christian skew to the right in 2016.

As these five images scroll though we can see some interesting conclusions:


- As we have seen earlier, as wealth increases, support for Labour decreases.


- Alternately, as wealth increases support for both the Coalition and the Greens also increases.


- In the most deprived areas, support for Labour still decreases as the percentage of Christian's increases, with the vote moving to the Coalition.


- In all other (lesser deprived) areas, support for Labour is not impacted by religious affiliation.  It could in fact be argued for high levels of wealth, support for Labour marginally increases as the level of Christianity increases.


- In places of high wealth, the religious skew tends to drive preference between the Greens and the Coalition - areas of low religion have a high preference for the Greens, whereas the Coalition is at its most dominant amongst more wealthy Christian areas.


So what does this tell us about the Christians?  They certainly lean to the right, but interestingly this is mostly at the extremes of wealth, the top 20% and the bottom 20%.  The Christian vote for Labour more or less reflects socio-economic expectations (except within the most deprived) but very little interest in the Greens.


And what can political parties learn from this, especially given voters who affiliate as Christians are more than half of the electorate?  The Coalition clearly has a message that has affinity with the Christian population although is perhaps not as clear to the middle class.  Labour is not engaging as well and would in particular be concerned it is not reaching its core base of the most deprived amongst Christians.  The Greens pick up most of their support from the upper-middle class, but certainly not amongst Christians. 


The size of the data means it is quite difficult to do this same analysis for the other main religions.  The other group we can look at are those who explicitly state they have no religion (which was 22% of the population in 2011).  This can be seen in aggregate below.

As we would expect this roughly the opposite profile of the Christian group, although at more extreme levels.  This is not unusual as this is an extreme position.  We might expect similar if we only looked at church-going Christians in these statistics (instead of just Christian affiliation).   It is clear that the Green party picks up a large percentage of its support from the no religion group.  The fact that Labour makes little in roads amongst this group suggests are they much more (relatively) popular amongst other religions.


Again we can see this same chart for just the capital cities at the five levels of wealth.

Again we have some interesting findings:


- Green support is concentrated amongst the top 40% of wealth.


- Labour struggles to connect the no religion group at all levels of wealth.


- Both One Nation and the Xenophon Team are very popular amongst people of no religion and with high levels of deprivation.


In summary, although we can't isolate with precision the relationship between religion and voting preference we can still draw some very solid conclusions.  It again shows what can be done with data sources which are freely available with a little bit of creativity.  Although we can accounted for, as best we can, the impacts of geography and wealth, other biases will still exist.  We have not considered age which would also have some swing, particularly given the Christian population is likely to be older than average.  Nonetheless, this insight is significant for both political groups and religious groups as they look to see how well they engage with different parts of society, and more importantly, what they can do about improving their engagement.


We can look forward to the release of 2016 census religion data which will enhance this work considerably.  Firstly it will be more up to date.  But more significantly we can expect to see a rising level of no religion and a reduced percentage of Christians.  This Christian group is also likely to be more distinctive from the overall population (compared to 2011) which will without doubt make the findings more fascinating.  Watch this space.