Election 2017: Counting on the Maps
This post looks at mapping the results of the 2017 general election with a particular focus on using cartograms to better represent the results.
I don’t know about you, but on election night I was looking for some maps. Sure, a ticker tape of the electorate outcomes across the bottom of the screen was useful, and the talking heads were in fine form, but the maps were few and far between. I managed to find a map on the Radio New Zealand website which presented the results of the electorate vote but couldn’t find it the next day.
TVNZ managed to make one available and here it is:
Interestingly enough, there are sort of two maps here – the map on the left showing the election results by electorate and the semi-map at right showing how this translated to seats in Parliament. The main map is an “interactive” map in the sense that you can click on it and get some more details, in this case, the results for the electorate you clicked on:
So somewhere, a server has the GIS data on electorate boundaries and an attribute table that holds the details and is streaming these data over the web. My concern here is that this kind of map often distorts the results. Let’s look at the main map in detail:
First off, the boundaries look at bit puffy at the edges. This is because the electorate boundaries include the 12 mile offshore territorial limit. Second, and more importantly, looking at this map one could be forgiven for thinking that the National party won in a landslide. I mean look at all that blue!
It’s been long recognised that mapping things like elections this way can give a skewed interpretation of results, mainly because areas tend to be coloured one way or another based on the results regardless of population. In the absence of any information we might naturally assume that these are all areas of equal population and therefore with equal impact. Most certainly tend to interpret it this way. Looking at the map above, the population of the West Coast electorate is 31,248 according to Stats New Zealand so, given its size, its effect on the vote must have been quite significant. But compare that with the three Christchurch electorates that went to Labour – their combined population is 124,680! So while these maps are correct in their spatial extent they are misleading in this context because they don’t adequately convey the impact of population differences. But if we’re willing to sacrifice spatial accuracy, there is another approach: cartograms.
With Cartograms, we can distort areas to represent some other variable – often its population but it could be anything such as CO2 production or the number of broadband connections. For instance, below is an image I scanned in from my Bateman Contemporary Atlas of New Zealand (2004). It shows a cartogram of NZ based on population:
There’s enough spatial familiarity that we can know roughly what’s where and while it’s no longer spatially correct, it’s conveying another dimension of the place we live (“It’s a map, Jim, but not as we know it.”) Another quick example going back to the 2016 US election – here’s the standard result map – red for Republicans, blue for Democrats:
And here’s a cartogram based on population:
Despite continually being told by someone who will not be mentioned in these pages how big the election win was, this puts a slightly different perspective on it I’d say.
So I set out to do something similar with our recent election results – I’ll give you a quick summary of how I set it up and then you can be the judge if it’s any more enlightening. All the data I used can be found in J:\Current_Projects\Blog\2017Election if you want to have a look.
First off, I needed the electorate boundaries. From Koordinates.com I got the 2014 General Electorate Boundaries, which were used for this election (these could also be found on the StatsNZ website). While I was there I also downloaded the Maori Electorate Boundaries. Here are the general electorates:
With this tool I could search for a particular set of data (population by electorates in this case) and download the results as a CSV or Excel spreadsheet. The initial sheet had the data I needed:
Conveniently, the electorate names match those I have in my spatial layer. But these data needed to be reformatted before I could use them in ArcMap:
NB: these are total populations rather than voter eligible populations but I hope you get the general idea. With a table join, I could add the populations to my data layer and display them as classes:
Next I had to add an attribute that held which party won each electorate. I couldn’t find a spreadsheet or file with this information so rather than type in the party record by record, I first set them all to National (don’t read too much into that, Bill…) and then went through and selected all the ones I knew went to Labour and did a quick field calculation. The last one to deal with was Espom which went to ACT. Here’s what this result looked like, echoing back to the first map we saw above.
So all I’ve really done here is recreate the TVNZ map – the next step is where it gets interesting. There’s a tool for ArcGIS that can be downloaded and used to create your own cartograms (using the Gastner-Newman method). With a bit of fiddling, here, ta da, is my result:
Quite different, eh? Here the areas have been reshaped based on population and coloured by party. Epsom, which wasn’t even visible before, now stands out a bit more. And while National did win most of the rural electorates, they are now a bit more balanced by the mostly urban areas that Labour won. When shown this way, I would argue that it provides a bit more context to the geographically correct map. What do you think?
NB: These maps were created before all the special votes were counted so the final results may change.
Here’s how the Maori Electorates look after being cartogrammed:
All red as they all went to Labour but conveying a sense of how populations are distributed. Stephen Beban presents another cartogram alternative, the “hexamap” which shows each electorate as a set of five hexagons:
So cartograms offer another way of presenting these results in a way that tries to balance out the over- and understatements inherent in a standard map. Another interesting one I came across maps all the voting stations and breaks them down by results. Maps have a lot to tell us about elections but sometimes being spatially correct doesn’t tell the story as well. (If you want another classic example, check here).