Inner City Voting in 2022

With an imminent state election, I thought it’d be a nice little mapping exercise to look at Greens support from the previous state election in some inner-city electorates (formally known as “Districts”). In particular: Brunswick, Footscray, Melbourne (current held by the Leader of the Victorian Greens), Northcote, Pascoe Vale, Prahran, Preston, Richmond. These are the seats in which the Greens were “two candidate preferred” (2cp); so that in the final round of counting, it was Greens against another party. This means that when all but two of the candidates have been eliminated, one of the final two was a Greens candidate. (Note that Australia uses Instant Run-Off Voting.)

In the map, each district is subdivided into regions about each polling booth. In particular, for each each region, each point in it is closer to its booth than to any other booth. In other words, each region contains those points which are closest to its booth. (This sort of subdivision is called a Voronoi Diagram.)

I have made some assumptions to make these maps, so as to include all votes. The VEC publishes data from all elections: for the 2022 state election there are results for every district. Such results include the first preference votes for each candidate by voting centre, as well as the 2cp results after all votes had been distributed according to IRV. However, we also need to take into account early votes and postal votes: these can amount to a similar number to the votes cast on polling day. However (obviously) it makes no sense to talk about early votes and postal votes by voting centre.

What I did was distribute the 2cp results from the early and postal votes into the districts by weights given by polling numbers on the day. Suppose there are \(n\) voting centres, and the 2cp results for the final candidates Alice and Bob at the \(i\) -th centre are \(a_i\) and \(b_i\) respectively. Let the sums be

\[ A = \sum_i a_i,\qquad B = \sum_i b_i. \]

Suppose also that the 2cp results from the early and postal votes sum to \(\alpha\) and \(\beta\). These values are then added to the 2cp values at each centre to give

\[ \alpha_i = a_i(1+\alpha/A),\qquad \beta_i = b_i(1 + \beta/B). \]

We can see that this works:

\[\begin{align*} \sum \alpha_i&=\sum a_i(1+\alpha/A)\\ & = (1+\alpha/A)\sum a_i\\ & = (1 + \alpha/A)A\\ & = A + \alpha \end{align*}\]

Thus the sum of all 2cp values as given by our formula is the sum of polled votes, and early and postal votes.

What all this means is that I’ve assumed:

  • in general, people will vote at the polling booth closest to their homes

  • people who voted early, or by post, would have voted at their nearest booth if they had cast a ballot on the day.

With these assumptions, we can build a map of Greens support over a district.

In a map, you can click on the point representing its region to see the booth name, and the total votes for the two final candidates.

The map: