Investigation into application-based positioning data for COVID-19
As part of its COVID-19 response, the ABS investigated application-based geospatial positioning data from mobile phones as a source of information on social and economic changes during the coronavirus crisis.
Many ad-supported mobile applications (‘apps’) use Global Positioning System (GPS) data to serve location-specific advertising, if users have given explicit permission for the apps to access that positioning data. In recent years researchers have started using app-sourced location data to explore social issues including economic and epidemiological aspects of the COVID-19 crisis.
In 2020 the ABS purchased two samples of app-based location data from a commercial provider, covering the state of Victoria for January-April 2020. These samples had been aggregated to Statistical Area Level 1 (SA1) and whole-of-state level. No information about individuals was obtained, nor any information that could be used to identify individuals. The data is not related in any way to the Australian Government’s COVIDSafe app.
The first sample describes ‘close contact’ events where two mobile devices are observed within five minutes and five metres of one another (five metres being the limit of GPS precision). One version of the aggregated data assigns contacts to the regions where they occur (‘event-location’); the other assigns them to the ‘common evening locations’ of the devices involved.
Comparison to publicly available sources of data (Apple and Google mobility data, City of Melbourne pedestrian counters) gave broadly consistent results, with all sources showing a sharp drop-off in activity from mid-March 2020, but the ‘close contacts’ data has two major advantages.
First, it allows for state- or nation-wide coverage at fine levels of geography, something not available from the public sources.
Second, the ability to view both ‘event-location’ and ‘common evening location’ data offers insight into patterns of activity. For instance, contacts may increase within a residential suburb during lockdown, because residents who previously commuted elsewhere during the day are now spending more time in their home suburb. Examining the “common evening location” view helps understand this pattern: residents are making fewer contacts overall and these contacts have moved from the city to the suburb.
The second sample gives ‘foot traffic’ data: daily counts of devices observed at various locations of interest. These locations are primarily retail stores for major brands, but also include some other location types such as universities, Centrelink offices, and national parks. Comparison with other sources available to ABS showed that for many categories of retail business, foot traffic tracked closely with business income.
The ABS is now considering further applications for this type of data.
For further information, please contact Geoffrey Brent at email@example.com.