UPDATE: Labor wins in a landslide

A swing of 10.1% away from the Coalition was required for the Labor Party to win the 2017 State election.  While the Labor Party enjoyed a 9.1% swing to its favour, the critical factor - a surprise to many - was an even larger swing of 15.9% away from the Liberal Party.  The Coalition lost voters not just to Labor, but to One Nation and other minor parties, with Labor then preferenced higher than the Coalition, ultimately costing them the election.  A key example of this is the seat of Murray-Wellington.  Here Labor had a 7.9% swing towards them, but the Liberals had a massive 20.5% swing against, fuelled by the new candidates from One Nation and Shooters, Fishers, Farmers picking up 11.3% and 6.4% of the first preferences respectively.

The cartograms of the election results show that the metropolitan fringe previously held by Labor has extended well into the central suburbs, and the Coalition only has three “safe” seats remaining – Churchlands, Cottesloe and Vasse.  Labor now has 24 “safe” seats, and with just 30 required to win, the Coalition has a tough job ahead of them come the next election.

Election results: http://www.abc.net.au/news/elections/wa-election-2017/
Marginal/Safe seat definitions: http://www.aec.gov.au/footer/Glossary.htm

March 2017

WA Election 2017 - Where will it be won or lost?

Visualisation using Cartograms

Lots of statistics get tossed about during elections, sometimes to inform and sometimes, it seems, to confuse.  One of the challenges when presenting statistical information to a wide audience is how to communicate the diversity between the electorates – the election result depends upon the collection of outcomes in individual electorates.  In the case of Western Australia this is compounded by the enormous differences in the physical sizes of electorates and, to a lesser extent, the differences in their populations.

Here we are focusing on the Western Australian Legislative Assembly, where, after 11th March, the next government will be formed.  The Assembly has 59 seats.  The largest electorate is the North West Central at 815,387 square kilometres; the smallest is Mount Lawley at 18 square kilometres.  In terms of electors, the largest is Swan Hills at 26,987, the smallest North West Central at 10,510.

Electorates are geographical areas and hence it is natural to try to display them as a map, as on the right.  However the disparity in size means that a map that covers the entire state is unable to show the detail of the South-West, and a map of the South-West does not show the full detail of the metropolitan area.  Indeed the map of the whole state suggests that the electors overwhelmingly voted for the National Party at the last election, when in fact they hold just 7 of the 59 seats.

Clearly a better display is needed.

A traditional map is usually one that approximately represents physical areas.  That is, if two regions have the same physical area then they will have close to the same area on the map.  

Cartograms are maps where the area on the map represents something of interest.  A common form of cartograms uses areas to represent population so that equal areas on the map represent equal populations.  There are various methods of producing cartograms but they usually start with a standard map and distort it to achieve their end.  The mathematics of this distortion process is complex and the algorithms need to be applied with care, but the results can be fascinating.

Since the election result is determined by the numbers of members in the Assembly and each electorate returns just one member, it is logical to use a cartogram that gives each electorate the same size, so that areas on the map will correspond to numbers in the Assembly.  The cartogram is geographically highly distorted, but it immediately shows that the election is largely decided in the metropolitan area simply because that is where most of the electorates are.  The apparent dominance of the Nationals disappears and the Liberals are clearly seen as the majority in the current Assembly.  

The clusters of which party holds each seat is clear in the cartograms.  Labor is dominant in the south-west and north-east metropolitan, while the Liberals are dominant throughout most of the rest of the metropolitan area.  The Nationals are somewhat unkindly reduced to a fringe.  The exceptions are the Kimberley and Albany held by Labor and Geraldton held by the Liberals.

What will happen at the March election?  Here we will not go into the details of swings (made particularly complex by the recent entry of One Nation fielding candidates in 35 electorates) so we defer to the excellent modelling of Antony Green on his ABC website.  He estimates that a 10.1% swing in the two party preferred vote is required for Labor to achieve government and for that swing has predicted the outcome in each of the 59 seats. 

The difference in the cartograms (of current seat holder versus after the swing to Labor) shows that the swing, if it occurs, will largely be in the north-east metropolitan area, joining up the seats already held there into a single group.

Voting behaviour is determined by many issues.  Some of these have a strong geographic basis and it is interesting to see how these vary across electorates.

One such issue affecting voting behaviour is the socio-economic advantage (or disadvantage) of the electors.  The Australian Bureau of Statistics (ABS) index of advantage-disadvantage is one standard way of representing economic and work status, based on the 2011 Census.  High numbers represent advantaged areas and low numbers disadvantaged areas, with 1000 representing the Australian average.  Broadly it can be seen that electors in areas of higher socio-economic advantage are more likely to vote Liberal.  The sharp divide between the metropolitan and non-metropolitan electorates is also evident.

Population change and the economic pressures of developing suburbs – the “mortgage belt” - are also relevant.  Again the ABS provides data on this.  The population growth is occurring in a ring around the metropolitan area, generally affecting seats that are or might be Labor.

While cartograms are not an ideal solution to all information display issues, they demonstrate the value of thinking outside the box when it comes to presenting data, whether it be geographical (maps) or other sorts of data.  Data Analysis Australia is always on the lookout for non-standard ways to present results that can provide illuminating insight not evident via conventional displays.


  1. WA Electoral Boundaries (EB) - the 2017 electoral boundaries were finalised in 2015 (after re-distribution) and will be used for the March 2017 election. Geographic data was sourced from the Office of the Electoral Distribution Commissioners.
  2. Cartogram Style - The cartograms are created using the free JAVA tool Scape Toad and use the Gastner and Newman 2004 algorithm.
  3. Incumbent Party - this information came from the Parliament WA website.
  4. Socio-Economic Index  - is from the Socio-Economic Indexes for Areas (SEIFA) as determined from ABS 2011 Census data. 
  5. Regional population change 2014-2015 - ABS Catalogue 3218.0 Regional Population Growth, Australia. 

February 2017