Predictors of Australian Election Results
Authors: Lachlan Thomas Moody
Abstract: The 2019 Australian federal election was one of the most unexpected in recent memory, which Liberal Party presentative Scott Morrison defying the predictions of the media and pollsters to become Australia's next Prime Minister. This interactive visualisation dashboard will explore in three parts why this was the case and what factors contributed to his victory.
Full Description: This visualisation project, titled Predictors of Australian Election Results, was completed as part of a university assignment and designed to explore the results of the 2019 Australian Federal election and whether the distribution of various demographic statistics could be modelled and used to predict outcomes at the electorate level. In order to formally address this topic, the following three research
questions were proposed:
-
How were Australia’s national election results distributed in 2019 on a two-party preferred basis?
-
How were Australia’s national demographic statistics distributed in 2019 for each electorate?
-
Is there any interaction or explanatory power between demographic statistics and election results in Australia?
With the above research questions and findings in mind, several key messages have been intentionally conveyed by the final visualisation.
Relating to the first question, users are able to identify several important characteristics of Australia’s election results and how they were distributed across the country. This
includes that the Liberal Party won the election mainly due to strong results in Queensland and Western Australia and also, the average size of the electorates won by the two parties was drastically different with the Liberal Party securing many seats in large regional areas while Labor performed better in metropolitan centres.
Secondly, question two related to the distribution of various demographic variables
amongst these 151 electorates. Due to the interactive nature of the designed visualisation, all of the demographic variables can be viewed by users on a map of the country. In addition, the distribution and averages of these variables nationally and within Liberal and Labor electorates visually can be seen. While more exploratory in nature, users can select the variables that are the most interesting or important to them to gain a better understanding of how they are represented across Australia. For example, for the Voter Age statistic, users can see that nationally the average is 47 whereas for Liberal electorates it is 48.21 and for Labor ones it is 45.64. This helps convey to the user the difference in demographic trends between seats
that lean towards Liberal or Labor. Similar findings can be deduced for all other variables.
The last research question is based on a combination of the two preceding ones as it relates to a possible interaction between the two. The inclusion of a visualisation relating to this conveys to users that election results and demographic statistics are indeed related and that these different variables can quite accurately predict the outcome of the election. The inclusion of several input sliders in the design illustrates which variables are of particular importance in modelling the outcome. Additionally, as users interact with the sliders, they are able to see precisely how it affects the predicted outcome. For example, if Voter Age (discussed above) were to decrease to 46, only 1 year below the current average, the predicted outcome would flip in the Labor Party’s favour. This also helps covey to audiences the message about which variables are more important to each party’s performance in an election.
Finally, regarding the intended audience of the narrative visualisation, as election results are off interest and impact on everybody in the country, the design was created for the Australian general public. As such, in general basic and easy-to-understand plots and graphics were chosen including choropleths, bar charts and histograms.
Keywords: politics, election, spatial, leaflet, ggplot, modelling, interactive, exploration, dashboard
Shiny app: https://ltmoo2.shinyapps.io/lachlanmoody_27809951_code/
Repo: GitHub - ltmoo2/Shiny-Contest-Submission
RStudio Cloud: RStudio Cloud
Thumbnail:
Full image: