Pictures that tell us what we already knew/suspected.
You nerds seemed to like these charts, so I thought I’d write something a bit longer to cover off: how I did it, what it shows, and the 6-or-so grains of salt it needs to be taken with.
Method
Each of these heat maps shows the relationship between: the percentage of the party vote each party received at each polling place in 2014, and the average age (for the total population, not just eligible voters) and average household income of the StatsNZ Area Unit (read: suburb) the polling place is located in, according to the 2013 Census.
Each of the 100 shows the percentage of the vote the party received at the polling places which meet the relevant age and income criteria
The age and income data was aggregated into deciles (relative to the area units which had polling places, not all area units), then matrix-ed out on a 10x10 grid. The vote-side data is fairly self explanatory.
The Excel file used to do all this can be found here.
The underlying data comes from a database I’ve rigged up using information published by the Electoral Commission. I’ll publish the full thing soon, but I need to tidy it up a bit first.
Comments
Might as well say it: the older and the richer your area, the more likely you are to vote National, and the less likely you are to vote Labour. At the same time, even in their weakest spot, National still score 22%, which indicates just how favourable the ‘baseline’ is to them.
Other fun points, both age (older) and income (poorer) correlate with supporting NZ First, but income is a stronger driver than age.
About that concentration of the Green’s in the bottom right… This is not saying most of the Greens vote comes from the young and rich (see health-warnings below). What it’s saying is that the Greens do comparatively better in areas where the population is on average both young and richer.
The specific places captured in those boxes are for the most part Wellington Central and Auckland’s inner west (Grey Lynn, Kingsland). Now, as I’ll show in a future post, there’s a lot of Green vote there, but on the other hand there’s still a lot (a majority, nearly) of Tories too.
Health Warnings
- It’s about place, not people
I did the analysis at an area-unit and polling place level because that’s as granular as you can get. But I can’t stress this enough, this is a guide only. All the polling places and suburbs here have a lot of internal diversity, so don’t make too much of the trend. As a helpful tweep pointed put, there’s a lot of ecological fallacy risk here
2. Not all squares are created equal
Unsurprisingly, age and income are not independent variables. On top of that, the scatter of polling places and votes between the boxes is not even.Note the 10:1 ratio between the biggest and smallest units.
3. It’s where people vote, not where they live
The whole game relies on the assumption that the voters voting in a given booth reflect (on average) the characteristics of the area around the booth. A reasonable assumption, I think (not sure how we’d test that, open to ideas). As the map below shows, the pattern of area units and polling booths don’t map 1:1.
In sum
This was basically done on a lunch break by an ex-law student who took an intro to Excel course one time, so bear that in mind before treating this a gospel.
Self-depreciation aside, the patterns, especially for National and Labour, are pretty clear and consistent. I’m looking forward to seeing if/how they shift after September 23.
