Federal election data published by the Australian Electoral Commission is excellent. Polling place-level results are published in easy-to-analyse CSV files, and can be easily manipulated to give you whatever figures you need. Full sets of polling place address data, including latitudes and longitudes, are also easily accessible, along with the candidate lists.
The same is, sadly, not true for most state and local elections.
For a while now I’ve spent a great deal of energy pulling out inaccessible election results, published as HTML tables at the seat level, as PDFs or god forbid as PNG image files (hello Tasmanian Electoral Commission). I’ve now started a gradual project of taking this data as I need it, and turning it into standardised and complete datasets.
Each dataset includes:
- Full list of candidates, including ballot order, party affiliation and group where applicable.
- Full list of polling places, including latitude and longitude, and including IDs to match the polling place address data to voting data.
- Full set of vote data at the polling place and seat/ward level broken down by candidate. In some cases, this is also available at the two-party-preferred and/or two-candidate-preferred level.
The following is the list of all elections for which this task has been completed. I’ll post occasionally when I complete new datasets to add to this collection. Click on each link to go to the Google Drive folder containing the data:
- Australian Capital Territory 2012
- Brisbane City Council 2012
- NSW council elections 2011 – special elections were held for Wollongong and Shellharbour councils
- NSW council elections 2012
- NSW council elections 2016 – elections were only held in councils unaffected by planned amalgamations. This data includes the delayed Tweed council election.
- Northern Territory 2012
- Queensland 2015
- Tasmanian Legislative Council 2007-2016
- Western Australia 2013
Thanks for help
I’d particularly like to thank Alex Lum who managed to pull out the links to all of the NSW 2016 council election vote preferences, which made it possible to convert them into booth-level data.