Friday, 31 July 2015

Party Vote Demographics: Part III

In this series of blog posts, I’m looking at electorate demographics from the 2013 census, and the relationship with votes for political parties in the 2014 election. This is an important disclaimer, so I will repeat it again: correlation does not imply causation. This post will look at religion, marriage, and some other claims.

“Christians vote Conservative and United Future”
In electorates where there are more people who declare a religious affiliation with …
… Christianity (all denominations) (Anglican, Baptist, Catholic, Christian nfd, Latter-day Saints, Methodist, Pentecostal, Presbyterian, Congregational, Reformed, and Other Christian), fewer people voted for the Greens (r=-0.568)
… Māori Christianity (all denominations) (Rātana, Ringatū, Other Māori Christian), a lot more people voted for InternetMANA (r=+0.829) and the Māori Party (r=+0.844)
… Baptism, more people voted Conservative (r=+0.469) and ACT (r=+0.337)
… Anglicanism, slightly more people voted Conservative (r=+0.150), and fewer people voted for Labour (r=-0.486)
… Catholicism, Latter-Day Saints, Methodists, Pentecostals, and Other Christians, more people voted for Labour (r≈+0.5)
… Judaism, more people voted for ACT (r=+0.426), slightly more people voted for the Civilian Party (r=+0.248), United Future (r=+0.221), and National (r=+0.197), and fewer people voted for NZ First (r=-0.466)
… Buddhism, a lot more people voted for ACT (r=+0.819), and fewer people voted for NZ First (r=-0.597) and ALCP (r=-0.657)
… Hinduism, more people voted for Labour (r=+0.565) and ACT (r=+0.448), and fewer people voted for ALCP (r=-0.574)
… Islam, more people voted for Labour (r=+0.603) and ACT (r=+0.446), and fewer people voted for ALCP (r=-0.587) and NZ First (r=-0.495)
… Spiritualism and New Age religions, slightly more people voted for the Greens (r=+0.146)
… no religion, fewer people voted for Labour (r=-0.313)

Discussion: As the number of non-religious people in New Zealand grows (from 29.6% in 2001 to 41.9% in 2013), the role of religion in politics should be gradually eroding away. However, we do have parties with strong ties to religious groups, such as the Conservative Party (apparently popular in electorates with more Baptists and Anglicans). There appears to be no relationship between United Future and number of religious people in a particular electorate, which is surprising given the party’s roots in the Christian-based Future New Zealand party.

There are probably some overlaps between religion and ethnicity; for example, electorates with more Māori Christianity more strongly supported the Māori Party and InternetMANA, which shouldn’t be too surprising. Religions more common among immigrants, including Buddhism, Hinduism, and Islam, followed similar patterns to correlations for recent immigrants. A few interesting observations can be made of the religious preferences of Labour and Green voting electorates, but I’m inclined to dismiss those correlations as coincidence mainly because I can’t think of any reason why those relationships should exist.

“[Insert political side here] are more likely to be married”
In electorates where more people are…
… married (not separated), more people voted for the Conservatives (r=+0.771), National (r=+0.760), and United Future (r=+0.331), and fewer people voted for InternetMANA (r=-0.505), the Māori Party (r=-0.455), and NZ First (r=-0.236)
… divorced, more people voted for NZ First (r=+0.637), the Conservatives (r=+0.273), and the Māori Party (r=+0.110)
… not (have never been) married or in a civil union, slightly more people voted for the Greens (r=+0.175), fewer people voted for NZ First (r=-0.236), the Conservatives (r=-0.752), and National (r=-0.760)

Discussion: There are claims on both sides; some people think right-wingers are more likely to be in marriages because of the religious connections, others think that left-wingers are more likely to be in marriages because of the family connections, and so on. The correlations suggest that electorates with more married people are more likely to vote right-wing, and that Māori Party and InternetMANA supporting electorates have fewer married people. Electorates with a lot of divorcees voted more for NZ First, and all the single ladies people electorates were marginally in more support of the Greens and very strongly not in support of the Conservatives or National. So maybe there could be some truth in the statement that single people are not right-wingers, until the day that they sign a bit of paper with a significant other so that the state recognises them as legally married, when they suddenly get hit by a wave of neoliberalism and realise that Colin Craig had it right all along. Maybe.

“Smart/Educated (not synonymous) people won’t vote for the Civilian Party”
In electorates where there are more people...
… with Bachelors degrees, more people voted for the Civilian Party (r≈+0.38)
… with Honours degrees, more people voted for the Civilian Party (r≈+0.52)
… with Masters degrees, more people voted for the Civilian Party (r≈+0.42)
… with Doctorate degrees, more people voted for the Civilian Party (r≈+0.52)
… with tertiary education degrees, more people voted for the Greens (r≈+0.7), ACT (r≈+0.5), and United Future (r≈+0.37), and fewer people voted for ALCP (r≈-0.4) and NZ First (r≈-0.57)

Discussion: The argument was that more educated people take politics more seriously and wouldn’t dare to waste their vote on a joke party. The evidence reasonably clearly contradicts this, perhaps suggesting that satire might be relatively high-brow or that more educated people may be more disenfranchised with the political system and don’t care about their vote as much. Electorates with more educated people also voted more for the Greens and ACT, and less for ALCP and NZ First, which could be explained by a larger number of highly educated people living in urban/city areas, and thus capturing similar voting patterns as those discovered based on ethnicity and immigration. I note that education is not necessarily a good proxy for “smart”.

“People who use the internet will vote for the InternetMANA Party”
In electorates where there are more households with…
… access to the internet, there were no relationships with party vote
… access to a fax machine, slightly fewer people voted for InternetMANA (r=-0.160)
… access to a fax machine, a lot more people voted for the Conservatives (r=+0.523) and the National Party (r=+0.602)

Discussion: Okay so maybe no one really said this seriously, but it was fun to check anyway. Keith Ng found this surprising, since he “assume[d] people with fax machines were steampunks, therefore more likely to vote Internet[MANA]”. The reasonably strong relationship between fax machines and Conservative/National was a little surprising, but then again, maybe only old rich people can afford to maintain their fax machines. All that oiling and lubricating required to get those things going. Nothing beats the feeling of slightly warm slightly waxed paper telling you how much more money you’ve made. Toasty.

One important demographic that you might be wondering about is gender. There were no correlations found between number of votes for a political party in an electorate and the number of males/females in an electorate. This is largely due to the fact that we’re dealing at this rather coarse granularity, and all of the electorates have a roughly 49/51 male/female split, meaning that there are few differences between electorates.

You might also wonder why Labour, and to a lesser degree National, do not feature in these statistics a lot. The correlations are good for picking out odd relationships that would not appear if all characteristics were randomly (normally) distributed, which are typically relationships that only exist for a few categories of a demographic. Since Labour and National have relatively broad appeal, they largely attract people from all demographic groups, enough that it makes it statistically difficult to extract a relationship.

That’s it for demographic voting correlations. There were a lot more that could have been discussed, but a lot were quite specific (for example, 60-64 year old females who speak Tagalog) and probably not all that useful to discuss at such a broad level. Interesting? Entirely useless? How you interpret the statistics is ultimately up to you.

Okay, if you wanted some more, there's an appendix with some other pointless(ly fun) correlations.

But wait, there's more! Now you can see the data yourself, and pick your choice of political party and demographic variable - I made a visualisation, available here: andrewtzchendata.pythonanywhere.com (Note that it is a work in progress, so if it is completely broken when you try to look at it, try again in a few minutes).

No comments:

Post a comment