Oz Blog News Commentary

How much does the vote vary in each seat?

April 28, 2022 - 09:30 -- Admin

I was wondering recently about voting diversity within an electorate. I’m specifically talking about how much the vote varies from one booth to another across a local electorate’s geographic area. You certainly notice variation when looking at booth maps – some seats are very consistent, while others have some major differences between areas. But what happens if we try and quantify that variation?

I’ve found that, generally speaking, electorates covering larger land masses tend to have more variation in results, but there are some interesting exceptions.

For this blog post, I’m just focusing on two-party-preferred vote, and just looking at ordinary election day booths.

I started by simply measuring the distance between the most pro-Coalition booth and the most pro-Labor booth in each electorate. The biggest gap was in Parkes, where Labor polled 72.5% of the two-party-preferred vote in Menindee, and just 3.9% in Croppa Creek.

The gap is over 60% in seven other seats: Page, Eden-Monaro, O’Connor, Durack, Wide Bay, Macquarie and Kennedy.

At the other end of the list, the gap was just 18.5 percentage points in Bowman. Labor’s two-party-preferred vote varied from 37.3% to 55.8%.

Overall the seats with the biggest range tend to be big rural seats, but I think we could be a bit smarter in measuring this gap. There is a tendency in some seats like Parkes for there to be a small number of tiny booths which vote very differently to most of the seat. You can see a few specks of red on my Parkes booth map.

So for the next part, I cut out the 10% at either extreme (in terms of a proportion of the total formal votes cast at ordinary booths). This tends to remove the most extreme booths, particularly if they are small.

The gaps now get much smaller, and the list is not quite so dominated by big rural seats.

Macquarie tops the list, with a 41.0 percentage point gap between the 10th percentile and the 90th percentile. Labor polled 29.3% in Ebenezer, and 70.3% in Hazelbrook. If you check out the Macquarie seat guide, you can see a dramatic difference between the high Liberal vote in the Hawkesbury region and the high Labor vote in the Blue Mountains, particularly in the upper Mountains, and it’s not just a small extreme share of the vote.

This map shows this metric for all 151 seats:

While rural seats tend to have more variation, this isn’t always the case. Flynn, Parkes and Grey all have variation of more than 30 percentage points, but so does the south-western Sydney seat of Fowler, and the outer urban seats of Whitlam and Oxley.

One of the great things about this metric is you can then go visit a seat guide and see the map, which usually makes it clear why the seat has so much variation. The Fowler guide shows that most of the electorate is very strong for Labor, but there’s a handful of Liberal-voting booths at the north-western and south-eastern ends of the electorate. If you raised the threshold from 10% to 20%, you’d eliminate all those Liberal booths, while you wouldn’t change the upper extreme much.

I’ve made a chart showing the relationship between the size of an electorate and the variation in booth results. More accurately, I’m measuring the number of booths, which has a rough relationship with land mass, although some seats like Lingiari are massive but have very few booths due to a large share of the vote being cast through mobile voting teams.

!function(){"use strict";window.addEventListener("message",(function(e){if(void 0!["datawrapper-height"]){var t=document.querySelectorAll("iframe");for(var a in["datawrapper-height"])for(var r=0;r

While there is a relationship, there are exceptions, with Fowler and Macquarie particularly standing out.

At the other end of the spectrum, there is very little variation in the outer Brisbane seat of Petrie. The 10th percentile booth (Mango Hill) voted 39.6% for Labor, while the 90th percentile booth (Griffin) voted 48.1% for Labor.

Looking elsewhere on the map, it looks like there's more variation in urban Sydney seats than in Melbourne or Brisbane. Rural seats in southern NSW and Queensland seem to have less variation than in other rural areas.

I was particularly interested in Wentworth, a very small electorate which ranked tenth for the most variation across the middle 80% of the electorate. While the Liberal Party wins big along the harbourfront, with over 80% in two booths, Labor narrowly won the two-party-preferred vote in a bunch of suburbs closer to the beach in the Waverley area, along with a bunch of tiny booths at the CBD end of the seat.

Finally, I also tried another metric, called variance, which is defined as "a statistical measurement of the spread between numbers in a data set". I calculated variance just on the middle 80%. According to this metric, Macquarie is a clear leader, with Flynn, Fowler, Richmond and Wentworth rounding out the top five.

Bennelong has the least variance. There is a gap of just 9.2 percentage points between the 10th percentile and the 90th percentile. Labor barely won any booths in 2019, but the Liberal Party didn't exceed 60% of the two-party-preferred vote in many booths.

Hopefully this analysis can help you understand a bit more about different kinds of electorates - the sort of seat with strong contrasts between different parts of the seat will have a different kind of an election from a very homogenous seat.